Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+mmcv_full-1.5.0-cp310-cp310-linux_x86_64.whl filter=lfs diff=lfs merge=lfs -text

.github/workflows/lint.yaml

@@ -0,0 +1,39 @@
+name: Lint
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    branches:
+      - master
+      - 'gh/**'
+
+jobs:
+  run-linters:
+    name: Run linters
+    runs-on: ubuntu-20.04
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v3
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: 3.9
+          cache: 'pip'
+          cache-dependency-path: '**/requirements*.txt'
+      - name: Install Python (development) dependencies
+        run: |
+          pip install -r requirements-dev.txt
+      - name: Run flake8
+        run: |
+          flake8
+      - name: Run black
+        if: always()
+        run: |
+          black --check dinov2
+      - name: Run pylint
+        if: always()
+        run: |
+          pylint --exit-zero dinov2

.gitignore

@@ -0,0 +1,11 @@
+build/
+dist/
+*.egg-info/
+**/__pycache__/
+
+**/.ipynb_checkpoints
+**/.ipynb_checkpoints/**
+
+*.swp
+
+.vscode/

.vscode/launch.json

@@ -0,0 +1,16 @@
+{
+    // Use IntelliSense to learn about possible attributes.
+    // Hover to view descriptions of existing attributes.
+    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
+    "version": "0.2.0",
+    "configurations": [
+        {
+            "name": "Python: Current File",
+            "type": "python",
+            "request": "launch",
+            "program": "${file}",
+            "console": "integratedTerminal",
+            "justMyCode": false
+        }
+    ]
+}

CODE_OF_CONDUCT.md

@@ -0,0 +1,80 @@
+# Code of Conduct
+
+## Our Pledge
+
+In the interest of fostering an open and welcoming environment, we as
+contributors and maintainers pledge to make participation in our project and
+our community a harassment-free experience for everyone, regardless of age, body
+size, disability, ethnicity, sex characteristics, gender identity and expression,
+level of experience, education, socio-economic status, nationality, personal
+appearance, race, religion, or sexual identity and orientation.
+
+## Our Standards
+
+Examples of behavior that contributes to creating a positive environment
+include:
+
+* Using welcoming and inclusive language
+* Being respectful of differing viewpoints and experiences
+* Gracefully accepting constructive criticism
+* Focusing on what is best for the community
+* Showing empathy towards other community members
+
+Examples of unacceptable behavior by participants include:
+
+* The use of sexualized language or imagery and unwelcome sexual attention or
+advances
+* Trolling, insulting/derogatory comments, and personal or political attacks
+* Public or private harassment
+* Publishing others' private information, such as a physical or electronic
+address, without explicit permission
+* Other conduct which could reasonably be considered inappropriate in a
+professional setting
+
+## Our Responsibilities
+
+Project maintainers are responsible for clarifying the standards of acceptable
+behavior and are expected to take appropriate and fair corrective action in
+response to any instances of unacceptable behavior.
+
+Project maintainers have the right and responsibility to remove, edit, or
+reject comments, commits, code, wiki edits, issues, and other contributions
+that are not aligned to this Code of Conduct, or to ban temporarily or
+permanently any contributor for other behaviors that they deem inappropriate,
+threatening, offensive, or harmful.
+
+## Scope
+
+This Code of Conduct applies within all project spaces, and it also applies when
+an individual is representing the project or its community in public spaces.
+Examples of representing a project or community include using an official
+project e-mail address, posting via an official social media account, or acting
+as an appointed representative at an online or offline event. Representation of
+a project may be further defined and clarified by project maintainers.
+
+This Code of Conduct also applies outside the project spaces when there is a
+reasonable belief that an individual's behavior may have a negative impact on
+the project or its community.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported by contacting the project team at <opensource-conduct@meta.com>. All
+complaints will be reviewed and investigated and will result in a response that
+is deemed necessary and appropriate to the circumstances. The project team is
+obligated to maintain confidentiality with regard to the reporter of an incident.
+Further details of specific enforcement policies may be posted separately.
+
+Project maintainers who do not follow or enforce the Code of Conduct in good
+faith may face temporary or permanent repercussions as determined by other
+members of the project's leadership.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant][homepage], version 1.4,
+available at https://www.contributor-covenant.org/version/1/4/code-of-conduct.html
+
+[homepage]: https://www.contributor-covenant.org
+
+For answers to common questions about this code of conduct, see
+https://www.contributor-covenant.org/faq

CONTRIBUTING.md

@@ -0,0 +1,31 @@
+# Contributing to DINOv2
+We want to make contributing to this project as easy and transparent as
+possible.
+
+## Pull Requests
+We actively welcome your pull requests.
+
+1. Fork the repo and create your branch from `main`.
+2. If you've added code that should be tested, add tests.
+3. If you've changed APIs, update the documentation.
+4. Ensure the test suite passes.
+5. Make sure your code lints.
+6. If you haven't already, complete the Contributor License Agreement ("CLA").
+
+## Contributor License Agreement ("CLA")
+In order to accept your pull request, we need you to submit a CLA. You only need
+to do this once to work on any of Meta's open source projects.
+
+Complete your CLA here: <https://code.facebook.com/cla>
+
+## Issues
+We use GitHub issues to track public bugs. Please ensure your description is
+clear and has sufficient instructions to be able to reproduce the issue.
+
+Meta has a [bounty program](https://www.facebook.com/whitehat/) for the safe
+disclosure of security bugs. In those cases, please go through the process
+outlined on that page and do not file a public issue.
+
+## License
+By contributing to DINOv2, you agree that your contributions will be licensed
+under the LICENSE file in the root directory of this source tree.

Dockerfile

@@ -0,0 +1,11 @@
+FROM pytorch/pytorch:2.0.0-cuda11.7-cudnn8-devel
+
+COPY . /dinov2
+WORKDIR /dinov2
+
+RUN pip install -r requirements.txt
+RUN pip install -r requirements-extras.txt
+
+RUN apt-get update && apt-get install ffmpeg libsm6 libxext6  -y
+
+ 

LICENSE

@@ -0,0 +1,203 @@
+
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

MODEL_CARD.md

@@ -0,0 +1,201 @@
+# Model Card for DINOv2-S/B/L/g
+
+These are Vision Transformer models trained following the method described in the paper:
+"DINOv2: Learning Robust Visual Features without Supervision"
+
+We provide 4 models: 1 ViT-g trained from scratch, and 3 ViT-S/B/L models distilled from the ViT-g.
+
+## Model Details
+The model takes an image as input and returns a class token and patch tokens.
+
+The embedding dimension is: 
+- 384 for ViT-S.
+- 768 for ViT-B.
+- 1024 for ViT-L.
+- 1536 for ViT-g.
+
+The models follow a Transformer architecture, with a patch size of 14.
+
+For a 224x224 image, this results in 1 class token + 256 patch tokens.
+
+The models can accept larger images provided the image shapes are multiples of the patch size (14). 
+If this condition is not verified, the model will crop to the closest smaller multiple of the patch size.
+
+### Model Description
+
+- **Developed by:** Meta AI
+- **Model type:** Vision Transformer
+- **License:** Apache License 2.0
+
+- **Repository:** https://github.com/facebookresearch/dinov2
+- **Paper:** https://arxiv.org/abs/2304.07193
+- **Demo:** https://dinov2.metademolab.com/
+
+## Uses
+
+The models are vision backbones providing multi-purpose features for downstream tasks.
+
+### Direct Use
+
+The models can be used without fine-tuning, with downstream classifiers as simple as linear layers, to obtain competitive results:
+- on depth estimation, semantic segmentation, using linear layers.
+- on image classification, using k-NN classifiers on the class token.
+- on image classification, with logistic regression classifiers applied on the class token.
+- on image classification, with a linear layer applied on the class token and the average of the patch tokens.
+- on image retrieval using nearest neighbors.
+
+### Downstream Use
+
+It is technically possible to perform fine-tuning on the models, for small gains (we measured +2% on ImageNet-1k classification). 
+We recommend keeping this as a very last step and only when necessary, as the features already provide good performance out-of-the-box.
+
+## Bias, Risks, and Limitations
+
+Despite improvements thanks to the training method not using annotations, we still observe significant biases in our models toward rich households from Western countries.
+
+### Recommendations
+
+We expect fine-tuning will increase the biases in the features produced by the model as they will be tuned to the fine-tuning labels.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+```python
+import torch
+dinov2_vits14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vits14')
+dinov2_vitb14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitb14')
+dinov2_vitl14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitl14')
+dinov2_vitg14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitg14')
+```
+
+## Training Details
+
+### Training Data
+
+- **Training data:** LVD-142M (see paper)
+- **Training regime:** fp16 using PyTorch-FSDP mixed-precision.
+
+### Training Procedure 
+
+- **Training objective:**
+  - DINO self-distillation loss with multi-crop
+  - iBOT masked-image modeling loss
+  - KoLeo regularization on [CLS] tokens
+- **Architectures:**
+  - ViT-S (21M params): Patch size 14, embedding dimension 384, 6 heads, MLP FFN
+  - ViT-B (86M params): Patch size 14, embedding dimension 768, 12 heads, MLP FFN
+  - ViT-L (0.3B params): Patch size 14, embedding dimension 1024, 16 heads, MLP FFN
+  - ViT-g (1.1B params): Patch size 14, embedding dimension 1536, 24 heads, SwiGLU FFN
+- **Distillation:**
+  - Distillation follows the standard DINOv2 pretraining procedure, except the teacher is a pretrained ViT-g, frozen.
+
+## Evaluation
+
+We refer users to the associated paper for the evaluation protocols.
+
+<table>
+  <tr>
+    <th>model</th>
+    <th colspan="3">ImageNet-1k</th>
+    <th>NYU-Depth v2</th>
+    <th>SUN-RGBD</th>
+    <th>ADE20k</th>
+    <th>iNaturalist 2018</th>
+    <th>Oxford-H</th>
+  </tr>
+  <tr>
+    <th rowspan="2">task</th>
+    <th>classif. (acc)</th>
+    <th>classif. (acc)</th>
+    <th>classif. V2 (acc)</th>
+    <th>depth (RMSE)</th>
+    <th>depth (RMSE)</th>
+    <th>segm. (mAP)</th>
+    <th>classif. (acc)</th>
+    <th>retrieval (mAP)</th>
+  </tr>
+  <tr>
+    <!-- <th>^</th> -->
+    <th>k-NN</th>
+    <th>linear</th>
+    <th>linear</th>
+    <th>linear<br />4 layers</th>
+    <th>NYU-D transfer</th>
+    <th>multiscale</th>
+    <th>linear</th>
+    <th>nearest neighbor</th>
+  </tr>
+  <tr>
+    <td>ViT-S/14</td>
+    <td align="right">79.0%</td>
+    <td align="right">81.1%</td>
+    <td align="right">70.8%</td> 
+    <td align="right">0.417</td> 
+    <td align="right">0.431</td> 
+    <td align="right">47.2</td> 
+    <td align="right">69.5%</td> 
+    <td align="right">43.2</td> 
+  </tr>
+  <tr>
+    <td>ViT-B/14</td>
+    <td align="right">82.1%</td>
+    <td align="right">84.5%</td>
+    <td align="right">74.9%</td>
+    <td align="right">0.362</td> 
+    <td align="right">0.400</td> 
+    <td align="right">51.3</td> 
+    <td align="right">76.3%</td> 
+    <td align="right">49.5</td> 
+  </tr>
+  <tr>
+    <td>ViT-L/14</td>
+    <td align="right">83.5%</td>
+    <td align="right">86.3%</td>
+    <td align="right">77.6%</td>
+    <td align="right">0.333</td> 
+    <td align="right">0.396</td> 
+    <td align="right">53.1</td> 
+    <td align="right">79.8%</td> 
+    <td align="right">54.0</td> 
+  </tr>
+  <tr>
+    <td>ViT-g/14</td>
+    <td align="right">83.5%</td>
+    <td align="right">86.5%</td>
+    <td align="right">78.4%</td>
+    <td align="right">0.298</td> 
+    <td align="right">0.362</td> 
+    <td align="right">53.0</td> 
+    <td align="right">81.6%</td> 
+    <td align="right">52.3</td> 
+  </tr>
+</table>
+
+## Environmental Impact
+
+- **Hardware Type:** Nvidia A100
+- **Hours used:** 22,000 for ViT-g, 4,500 for ViT-S distillation, 5,300 for ViT-B distillation, 8,000 for ViT-L distillation
+- **Cloud Provider:** Private infra
+- **Compute Region:** USA
+- **Carbon Emitted:** 7t CO2eq
+
+#### Hardware
+
+Nvidia A100 GPUs
+
+#### Software
+
+PyTorch 2.0,
+xFormers 0.0.18
+
+**BibTeX**
+
+```
+@misc{oquab2023dinov2,
+  title={DINOv2: Learning Robust Visual Features without Supervision},
+  author={Oquab, Maxime and Darcet, Timothée and Moutakanni, Theo and Vo, Huy and Szafraniec, Marc and Khalidov, Vasil and Fernandez, Pierre and Haziza, Daniel and Massa, Francisco and El-Nouby, Alaaeldin and Howes, Russell and Huang, Po-Yao and Xu, Hu and Sharma, Vasu and Li, Shang-Wen and Galuba, Wojciech and Rabbat, Mike and Assran, Mido and Ballas, Nicolas and Synnaeve, Gabriel and Misra, Ishan and Jegou, Herve and Mairal, Julien and Labatut, Patrick and Joulin, Armand and Bojanowski, Piotr},
+  journal={arXiv:2304.07193},
+  year={2023}
+}
+```

README.md

@@ -1,12 +1,501 @@
 ---
-title: DinoV2 Semantic Segmentation
-emoji: 📊
-colorFrom: gray
-colorTo: pink
-sdk: gradio
-sdk_version: 3.44.1
+title: DinoV2_Semantic_Segmentation
 app_file: app.py
-pinned: false
+sdk: gradio
+sdk_version: 3.42.0
 ---
+# DINOv2: Learning Robust Visual Features without Supervision
+
+**[Meta AI Research, FAIR](https://ai.facebook.com/research/)**
+
+Maxime Oquab,
+Timothée Darcet,
+Théo Moutakanni,
+Huy V. Vo,
+Marc Szafraniec,
+Vasil Khalidov,
+Patrick Labatut,
+Armand Joulin,
+Piotr Bojanowski
+
+[[`Paper`](https://arxiv.org/abs/2304.07193)] [[`Blog`](https://ai.facebook.com/blog/dino-v2-computer-vision-self-supervised-learning/)] [[`Demo`](https://dinov2.metademolab.com)] [[`BibTeX`](#citing-dinov2)]
+
+PyTorch implementation and pretrained models for DINOv2. For details, see the paper: **[DINOv2: Learning Robust Visual Features without Supervision](https://arxiv.org/abs/2304.07193)**.
+
+DINOv2 models produce high-performance visual features that can be directly employed with classifiers as simple as linear layers on a variety of computer vision tasks; these visual features are robust and perform well across domains without any requirement for fine-tuning. The models were pretrained on a dataset of 142 M images without using any labels or annotations.
+
+https://github.com/facebookresearch/dinov2/assets/60359573/f168823e-7922-415a-b429-578badf5c356
+
+<div align="center">
+  Visualization of the three first principal components of the patch features of all frames, mapped to RGB values.
+</div>
+
+## Pretrained models
+
+<table style="margin: auto">
+  <thead>
+    <tr>
+      <th>model</th>
+      <th># of<br />params</th>
+      <th>ImageNet<br />k-NN</th>
+      <th>ImageNet<br />linear</th>
+      <th>download</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>ViT-S/14 distilled</td>
+      <td align="right">21 M</td>
+      <td align="right">79.0%</td>
+      <td align="right">81.1%</td>
+      <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_pretrain.pth">backbone only</a></td>
+    </tr>
+    <tr>
+      <td>ViT-B/14 distilled</td>
+      <td align="right">86 M</td>
+      <td align="right">82.1%</td>
+      <td align="right">84.5%</td>
+      <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_pretrain.pth">backbone only</a></td>
+    </tr>
+    <tr>
+      <td>ViT-L/14 distilled</td>
+      <td align="right">300 M</td>
+      <td align="right">83.5%</td>
+      <td align="right">86.3%</td>
+      <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_pretrain.pth">backbone only</a></td>
+    </tr>
+    <tr>
+      <td>ViT-g/14</td>
+      <td align="right">1,100 M</td>
+      <td align="right">83.5%</td>
+      <td align="right">86.5%</td>
+      <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_pretrain.pth">backbone only</a></td>
+    </tr>
+  </tbody>
+</table>
+
+### Pretrained backbones (via PyTorch Hub)
+
+Please follow the instructions [here](https://pytorch.org/get-started/locally/) to install PyTorch (the only required dependency for loading the model). Installing PyTorch with CUDA support is strongly recommended.
+
+A corresponding [model card](MODEL_CARD.md) is included in the repository.
+
+```python
+import torch
+
+dinov2_vits14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vits14')
+dinov2_vitb14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitb14')
+dinov2_vitl14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitl14')
+dinov2_vitg14 = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitg14')
+```
+
+### Pretrained heads - Image classification
+
+<table style="margin: auto">
+  <thead>
+    <tr>
+      <th rowspan="2">backbone</th>
+      <th>download</th>
+    </tr>
+    <tr>
+      <th>ImageNet</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>ViT-S/14 distilled</td>
+      <td>
+        linear head (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_linear4_head.pth">4 layers</a>)
+      </td>
+    </tr>
+    <tr>
+      <td>ViT-B/14 distilled</td>
+      <td>
+        linear head (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_linear4_head.pth">4 layers</a>)
+    </tr>
+    <tr>
+      <td>ViT-L/14 distilled</td>
+      <td>
+        linear head (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_linear4_head.pth">4 layers</a>)
+    </tr>
+    <tr>
+      <td>ViT-g/14</td>
+      <td>
+        linear head (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_linear4_head.pth">4 layers</a>)
+    </tr>
+  </tbody>
+</table>
+
+The (full) classifier models can be loaded via PyTorch Hub:
+
+```python
+import torch
+
+dinov2_vits14_lc = torch.hub.load('facebookresearch/dinov2', 'dinov2_vits14_lc')
+dinov2_vitb14_lc = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitb14_lc')
+dinov2_vitl14_lc = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitl14_lc')
+dinov2_vitg14_lc = torch.hub.load('facebookresearch/dinov2', 'dinov2_vitg14_lc')
+```
+
+### Pretrained heads - Depth estimation
+
+<table style="margin: auto">
+  <thead>
+    <tr>
+      <th rowspan="2">backbone</th>
+      <th colspan="2">download head</th>
+    </tr>
+    <tr>
+      <th>NYUd</th>
+      <th>KITTI</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>ViT-S/14 distilled</td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_nyu_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_nyu_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_nyu_dpt_head.pth">DPT</a>
+      </td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_kitti_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_kitti_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_kitti_dpt_head.pth">DPT</a>
+      </td>
+    </tr>
+    <tr>
+      <td>ViT-B/14 distilled</td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_nyu_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_nyu_dpt_head.pth">DPT</a>
+      </td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_kitti_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_kitti_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_kitti_dpt_head.pth">DPT</a>
+      </td>
+    </tr>
+    <tr>
+      <td>ViT-L/14 distilled</td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_nyu_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_nyu_dpt_head.pth">DPT</a>
+      </td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_kitti_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_kitti_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_kitti_dpt_head.pth">DPT</a>
+      </td>
+    </tr>
+    <tr>
+      <td>ViT-g/14</td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_nyu_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_nyu_dpt_head.pth">DPT</a>
+      </td>
+      <td>
+        linear (<a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_kitti_linear_head.pth">1 layer</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_kitti_linear4_head.pth">4 layers</a>),
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_kitti_dpt_head.pth">DPT</a>
+      </td>
+    </tr>
+  </tbody>
+</table>
+
+### Pretrained heads - Semantic segmentation
+
+<table style="margin: auto">
+  <thead>
+    <tr>
+      <th rowspan="2">backbone</th>
+      <th>download model</th>
+      <th colspan="2">download head</th>
+    </tr>
+    <tr>
+      <th>ADE20K</th>
+      <th>ADE20K</th>
+      <th>VOC2012</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>ViT-S/14 distilled</td>
+      <td></td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_ade20k_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_ade20k_ms_head.pth">multi-scale</a>
+      </td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_voc2012_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_voc2012_ms_head.pth">multi-scale</a>
+      </td>
+    </tr>
+    <tr>
+      <td>ViT-B/14 distilled</td>
+      <td></td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_ade20k_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_ade20k_ms_head.pth">multi-scale</a>
+      </td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_voc2012_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_voc2012_ms_head.pth">multi-scale</a>
+      </td>
+    </tr>
+    <tr>
+      <td>ViT-L/14 distilled</td>
+      <td></td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_ade20k_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_ade20k_ms_head.pth">multi-scale</a>
+      </td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_voc2012_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_voc2012_ms_head.pth">multi-scale</a>
+      </td>
+    </tr>
+    <tr>
+      <td>ViT-g/14</td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_ade20k_m2f.pth">Mask2Former</a>
+      </td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_ade20k_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_ade20k_ms_head.pth">multi-scale</a>
+      </td>
+      <td>
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_voc2012_linear_head.pth">linear</a>,
+        <a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_voc2012_ms_head.pth">multi-scale</a>
+      </td>
+    </tr>
+  </tbody>
+</table>
+
+## Installation
+
+The training and evaluation code requires PyTorch 2.0 and [xFormers](https://github.com/facebookresearch/xformers) 0.0.18 as well as a number of other 3rd party packages. Note that the code has only been tested with the specified versions and also expects a Linux environment. To setup all the required dependencies for training and evaluation, please follow the instructions below:
+
+*[conda](https://docs.conda.io/projects/conda/en/latest/user-guide/getting-started.html)* **(Recommended)** - Clone the repository and then create and activate a `dinov2` conda environment using the provided environment definition:
+
+```shell
+conda env create -f conda.yaml
+conda activate dinov2
+```
+
+*[pip](https://pip.pypa.io/en/stable/getting-started/)* - Clone the repository and then use the provided `requirements.txt` to install the dependencies:
+
+```shell
+pip install -r requirements.txt
+```
+
+For dense tasks (depth estimation and semantic segmentation), there are additional dependencies (specific versions of `mmcv` and `mmsegmentation`) which are captured in the `extras` dependency specifications:
+
+*[conda](https://docs.conda.io/projects/conda/en/latest/user-guide/getting-started.html)* **(Recommended)**:
+
+```shell
+conda env create -f conda-extras.yaml
+conda activate dinov2-extras
+```
+
+*[pip](https://pip.pypa.io/en/stable/getting-started/)*:
+
+```shell
+pip install -r requirements.txt -r requirements-extras.txt
+```
+
+## Data preparation
+
+### ImageNet-1k
+
+The root directory of the dataset should hold the following contents:
+
+- `<ROOT>/test/ILSVRC2012_test_00000001.JPEG`
+- `<ROOT>/test/[..]`
+- `<ROOT>/test/ILSVRC2012_test_00100000.JPEG`
+- `<ROOT>/train/n01440764/n01440764_10026.JPEG`
+- `<ROOT>/train/[...]`
+- `<ROOT>/train/n15075141/n15075141_9993.JPEG`
+- `<ROOT>/val/n01440764/ILSVRC2012_val_00000293.JPEG`
+- `<ROOT>/val/[...]`
+- `<ROOT>/val/n15075141/ILSVRC2012_val_00049174.JPEG`
+- `<ROOT>/labels.txt`
+
+The provided dataset implementation expects a few additional metadata files to be present under the extra directory:
+
+- `<EXTRA>/class-ids-TRAIN.npy`
+- `<EXTRA>/class-ids-VAL.npy`
+- `<EXTRA>/class-names-TRAIN.npy`
+- `<EXTRA>/class-names-VAL.npy`
+- `<EXTRA>/entries-TEST.npy`
+- `<EXTRA>/entries-TRAIN.npy`
+- `<EXTRA>/entries-VAL.npy`
+
+These metadata files can be generated (once) with the following lines of Python code:
+
+```python
+from dinov2.data.datasets import ImageNet
+
+for split in ImageNet.Split:
+    dataset = ImageNet(split=split, root="<ROOT>", extra="<EXTRA>")
+    dataset.dump_extra()
+```
+
+Note that the root and extra directories do not have to be distinct directories.
+
+### ImageNet-22k
+
+Please adapt the [dataset class](dinov2/data/datasets/image_net_22k.py) to match your local setup.
+
+<br />
+
+:warning: To execute the commands provided in the next sections for training and evaluation, the `dinov2` package should be included in the Python module search path, i.e. simply prefix the command to run with `PYTHONPATH=.`.
+
+## Training
+
+### Fast setup: training DINOv2 ViT-L/16 on ImageNet-1k
+
+Run DINOv2 training on 4 A100-80GB nodes (32 GPUs) in a SLURM cluster environment with submitit:
+
+```shell
+python dinov2/run/train/train.py \
+    --nodes 4 \
+    --config-file dinov2/configs/train/vitl16_short.yaml \
+    --output-dir <PATH/TO/OUTPUT/DIR> \
+    train.dataset_path=ImageNet:split=TRAIN:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET>
+```
+
+Training time is approximately 1 day and the resulting checkpoint should reach 81.6% on k-NN eval and 82.9% on linear eval.
+
+The training code saves the weights of the teacher in the `eval` folder every 12500 iterations for evaluation.
+
+### Long setup: training DINOv2 ViT-L/14 on ImageNet-22k
+
+Run DINOv2 training on 12 A100-80GB nodes (96 GPUs) in a SLURM cluster environment with submitit:
+
+```shell
+python dinov2/run/train/train.py \
+    --nodes 12 \
+    --config-file dinov2/configs/train/vitl14.yaml \
+    --output-dir <PATH/TO/OUTPUT/DIR> \
+    train.dataset_path=ImageNet22k:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET>
+```
+
+Training time is approximately 3.3 days and the resulting checkpoint should reach 82.0% on k-NN eval and 84.5% on linear eval.
+
+The training code saves the weights of the teacher in the `eval` folder every 12500 iterations for evaluation.
+
+
+## Evaluation
+
+The training code regularly saves the teacher weights. In order to evaluate the model, run the following evaluation on a single node:
+
+### k-NN classification on ImageNet-1k
+
+```shell
+python dinov2/run/eval/knn.py \
+    --config-file <PATH/TO/OUTPUT/DIR>/config.yaml \
+    --pretrained-weights <PATH/TO/OUTPUT/DIR>/eval/training_24999/teacher_checkpoint.pth \
+    --output-dir <PATH/TO/OUTPUT/DIR>/eval/training_24999/knn \
+    --train-dataset ImageNet:split=TRAIN:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET> \
+    --val-dataset ImageNet:split=VAL:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET>
+```
+
+### Logistic regression classification on ImageNet-1k
+
+```shell
+python dinov2/run/eval/log_regression.py \
+    --config-file <PATH/TO/OUTPUT/DIR>/config.yaml \
+    --pretrained-weights <PATH/TO/OUTPUT/DIR>/eval/training_24999/teacher_checkpoint.pth \
+    --output-dir <PATH/TO/OUTPUT/DIR>/eval/training_24999/logreg \
+    --train-dataset ImageNet:split=TRAIN:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET> \
+    --val-dataset ImageNet:split=VAL:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET>
+```
+
+### Linear classification with data augmentation on ImageNet-1k
+
+```shell
+python dinov2/run/eval/linear.py \
+    --config-file <PATH/TO/OUTPUT/DIR>/config.yaml \
+    --pretrained-weights <PATH/TO/OUTPUT/DIR>/eval/training_24999/teacher_checkpoint.pth \
+    --output-dir <PATH/TO/OUTPUT/DIR>/eval/training_24999/linear \
+    --train-dataset ImageNet:split=TRAIN:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET> \
+    --val-dataset ImageNet:split=VAL:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET>
+```
+
+We release the weights from evaluating the different models:
+
+<table style="margin: auto">
+  <tr>
+    <th>model</th>
+    <th>ImageNet<br />top-1</th>
+    <th>linear evaluation</th>
+  </tr>
+  <tr>
+    <td>ViT-S/14 distilled</td>
+    <td align="right">81.1%</td>
+    <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_linear_head.pth">linear head weights</a></td>
+  </tr>
+  <tr>
+    <td>ViT-B/14 distilled</td>
+    <td align="right">84.5%</td>
+    <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_linear_head.pth">linear head weights</a></td>
+  </tr>
+  <tr>
+    <td>ViT-L/14 distilled</td>
+    <td align="right">86.3%</td>
+    <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_linear_head.pth">linear head weights</a></td>
+  </tr>
+  <tr>
+    <td>ViT-g/14</td>
+    <td align="right">86.5%</td>
+    <td><a href="https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_linear_head.pth">linear head weights</a></td>
+  </tr>
+</table>
+
+The performance of the provided pretrained model weights can be evaluated as follows on ImageNet-1k:
+
+```shell
+python dinov2/run/eval/linear.py \
+    --config-file dinov2/configs/eval/vitg14_pretrain.yaml \
+    --pretrained-weights https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_pretrain.pth \
+    --train-dataset ImageNet:split=TRAIN:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET> \
+    --val-dataset ImageNet:split=VAL:root=<PATH/TO/DATASET>:extra=<PATH/TO/DATASET>
+```
+
+## Notebooks
+
+A few notebooks are provided to help the community leverage the models and code:
+
+<ul>
+  <li><a href="https://github.com/facebookresearch/dinov2/blob/main/notebooks/depth_estimation.ipynb">Depth estimation</a> - How to load and use the depth heads in combination with a matching backbone via mmcv</li>
+  <li><a href="https://github.com/facebookresearch/dinov2/blob/main/notebooks/semantic_segmentation.ipynb">Semantic segmentation</a> - How to load and use the segmentation heads in combination with a matching backbone via mmcv, and also how to load and use the Mask2Former-based segmentation model trained on ADE20K</li>
+</ul>
+
+## License
+
+DINOv2 code and model weights are released under the Apache License 2.0. See [LICENSE](LICENSE) for additional details.
+
+## Contributing
+
+See [contributing](CONTRIBUTING.md) and the [code of conduct](CODE_OF_CONDUCT.md).
+
+## Citing DINOv2
+
+If you find this repository useful, please consider giving a star :star: and citation :t-rex::
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+```
+@misc{oquab2023dinov2,
+  title={DINOv2: Learning Robust Visual Features without Supervision},
+  author={Oquab, Maxime and Darcet, Timothée and Moutakanni, Theo and Vo, Huy V. and Szafraniec, Marc and Khalidov, Vasil and Fernandez, Pierre and Haziza, Daniel and Massa, Francisco and El-Nouby, Alaaeldin and Howes, Russell and Huang, Po-Yao and Xu, Hu and Sharma, Vasu and Li, Shang-Wen and Galuba, Wojciech and Rabbat, Mike and Assran, Mido and Ballas, Nicolas and Synnaeve, Gabriel and Misra, Ishan and Jegou, Herve and Mairal, Julien and Labatut, Patrick and Joulin, Armand and Bojanowski, Piotr},
+  journal={arXiv:2304.07193},
+  year={2023}
+}
+```

app.py

@@ -0,0 +1,137 @@
+import os
+
+os.system("pip uninstall -y mmcv-full")
+os.system("pip uninstall -y mmsegmentation")
+os.system("pip install ./mmcv_full-1.5.0-cp310-cp310-linux_x86_64.whl")
+os.system("pip install -r requirements-extras.txt")
+# os.system("cp /home/user/data/dinov2_vitg14_ade20k_m2f.pth /home/user/.cache/torch/hub/checkpoints/dinov2_vitg14_ade20k_m2f.pth")
+
+import gradio as gr
+
+import base64
+import cv2
+import math
+import itertools
+from functools import partial
+from PIL import Image
+import numpy as np
+import pandas as pd
+
+import dinov2.eval.segmentation.utils.colormaps as colormaps
+
+import torch
+import torch.nn.functional as F
+from mmseg.apis import init_segmentor, inference_segmentor
+
+import dinov2.eval.segmentation.models
+import dinov2.eval.segmentation_m2f.models.segmentors
+
+import urllib
+
+import mmcv
+from mmcv.runner import load_checkpoint
+
+model = None
+model_loaded = False
+
+DINOV2_BASE_URL = "https://dl.fbaipublicfiles.com/dinov2"
+CONFIG_URL = f"{DINOV2_BASE_URL}/dinov2_vitg14/dinov2_vitg14_ade20k_m2f_config.py"
+CHECKPOINT_URL = f"{DINOV2_BASE_URL}/dinov2_vitg14/dinov2_vitg14_ade20k_m2f.pth"
+
+
+def load_config_from_url(url: str) -> str:
+    with urllib.request.urlopen(url) as f:
+        return f.read().decode()
+
+
+cfg_str = load_config_from_url(CONFIG_URL)
+cfg = mmcv.Config.fromstring(cfg_str, file_format=".py")
+
+
+DATASET_COLORMAPS = {
+    "ade20k": colormaps.ADE20K_COLORMAP,
+    "voc2012": colormaps.VOC2012_COLORMAP,
+}
+
+model = init_segmentor(cfg)
+load_checkpoint(model, CHECKPOINT_URL, map_location="cpu")
+model.cuda()
+model.eval()
+
+class CenterPadding(torch.nn.Module):
+    def __init__(self, multiple):
+        super().__init__()
+        self.multiple = multiple
+
+    def _get_pad(self, size):
+        new_size = math.ceil(size / self.multiple) * self.multiple
+        pad_size = new_size - size
+        pad_size_left = pad_size // 2
+        pad_size_right = pad_size - pad_size_left
+        return pad_size_left, pad_size_right
+
+    @torch.inference_mode()
+    def forward(self, x):
+        pads = list(itertools.chain.from_iterable(self._get_pad(m) for m in x.shape[:1:-1]))
+        output = F.pad(x, pads)
+        return output
+
+
+def create_segmenter(cfg, backbone_model):
+    model = init_segmentor(cfg)
+    model.backbone.forward = partial(
+        backbone_model.get_intermediate_layers,
+        n=cfg.model.backbone.out_indices,
+        reshape=True,
+    )
+    if hasattr(backbone_model, "patch_size"):
+        model.backbone.register_forward_pre_hook(lambda _, x: CenterPadding(backbone_model.patch_size)(x[0]))
+    model.init_weights()
+    return model
+
+
+def render_segmentation(segmentation_logits, dataset):
+    colormap = DATASET_COLORMAPS[dataset]
+    colormap_array = np.array(colormap, dtype=np.uint8)
+    segmentation_logits += 1
+    segmentation_values = colormap_array[segmentation_logits]
+    unique_labels = np.unique(segmentation_logits)
+
+    colormap_array = colormap_array[unique_labels]
+    df = pd.read_csv("labelmap.txt", sep="\t")
+
+    html_output = '<div style="display: flex; flex-wrap: wrap;">'
+    import matplotlib.pyplot as plt
+
+    for idx, color in enumerate(colormap_array):
+        color_box = np.zeros((20, 20, 3), dtype=np.uint8)
+        color_box[:, :] = color
+        color_box = cv2.cvtColor(color_box, cv2.COLOR_RGB2BGR)
+        _, img_data = cv2.imencode(".jpg", color_box)
+        img_base64 = base64.b64encode(img_data).decode("utf-8")
+        img_data_uri = f"data:image/jpg;base64,{img_base64}"
+        html_output += f'<div style="margin: 10px;"><img src="{img_data_uri}" /><p>{df.iloc[unique_labels[idx]-1]["Name"]}</p></div>'
+
+    html_output += "</div>"
+
+    return Image.fromarray(segmentation_values), html_output
+
+
+def predict(image_file):
+    array = np.array(image_file)[:, :, ::-1]  # BGR
+    segmentation_logits = inference_segmentor(model, array)[0]
+    segmented_image, html_output = render_segmentation(segmentation_logits, "ade20k")
+    return np.array(segmented_image), html_output
+
+description = "Gradio demo for Semantic segmentation. To use it, simply upload your image"
+
+demo = gr.Interface(
+    title="Semantic Segmentation - DinoV2",
+    fn=predict,
+    inputs=gr.inputs.Image(),
+    outputs=[gr.outputs.Image(type="numpy"), gr.outputs.HTML()],
+    examples=["example_1.jpg", "example_2.jpg"],
+    description=description,
+)
+
+demo.launch()

conda-extras.yaml

@@ -0,0 +1,24 @@
+name: dinov2-extras
+channels:
+  - defaults
+  - pytorch
+  - nvidia
+  - xformers
+  - conda-forge
+dependencies:
+  - python=3.9
+  - pytorch::pytorch=2.0.0
+  - pytorch::pytorch-cuda=11.7.0
+  - pytorch::torchvision=0.15.0
+  - omegaconf
+  - torchmetrics=0.10.3
+  - fvcore
+  - iopath
+  - xformers::xformers=0.0.18
+  - pip
+  - pip:
+    - git+https://github.com/facebookincubator/submitit
+    - --extra-index-url https://pypi.nvidia.com
+    - cuml-cu11
+    - mmcv-full==1.5.0
+    - mmsegmentation==0.27.0

conda.yaml

@@ -0,0 +1,22 @@
+name: dinov2
+channels:
+  - defaults
+  - pytorch
+  - nvidia
+  - xformers
+  - conda-forge
+dependencies:
+  - python=3.9
+  - pytorch::pytorch=2.0.0
+  - pytorch::pytorch-cuda=11.7.0
+  - pytorch::torchvision=0.15.0
+  - omegaconf
+  - torchmetrics=0.10.3
+  - fvcore
+  - iopath
+  - xformers::xformers=0.0.18
+  - pip
+  - pip:
+    - git+https://github.com/facebookincubator/submitit
+    - --extra-index-url https://pypi.nvidia.com
+    - cuml-cu11

demo.py

@@ -0,0 +1,153 @@
+import sys
+REPO_PATH="."
+sys.path.append("/dino_v2")
+
+import math
+import itertools
+from functools import partial
+
+import torch
+import torch.nn.functional as F
+from mmseg.apis import init_segmentor, inference_segmentor
+
+import dinov2.eval.segmentation.models
+
+
+class CenterPadding(torch.nn.Module):
+    def __init__(self, multiple):
+        super().__init__()
+        self.multiple = multiple
+
+    def _get_pad(self, size):
+        new_size = math.ceil(size / self.multiple) * self.multiple
+        pad_size = new_size - size
+        pad_size_left = pad_size // 2
+        pad_size_right = pad_size - pad_size_left
+        return pad_size_left, pad_size_right
+
+    @torch.inference_mode()
+    def forward(self, x):
+        pads = list(itertools.chain.from_iterable(self._get_pad(m) for m in x.shape[:1:-1]))
+        output = F.pad(x, pads)
+        return output
+
+
+def create_segmenter(cfg, backbone_model):
+    model = init_segmentor(cfg)
+    model.backbone.forward = partial(
+        backbone_model.get_intermediate_layers,
+        n=cfg.model.backbone.out_indices,
+        reshape=True,
+    )
+    if hasattr(backbone_model, "patch_size"):
+        model.backbone.register_forward_pre_hook(lambda _, x: CenterPadding(backbone_model.patch_size)(x[0]))
+    model.init_weights()
+    return model
+
+# BACKBONE_SIZE = "small" # in ("small", "base", "large" or "giant")
+
+
+# backbone_archs = {
+#     "small": "vits14",
+#     "base": "vitb14",
+#     "large": "vitl14",
+#     "giant": "vitg14",
+# }
+# backbone_arch = backbone_archs[BACKBONE_SIZE]
+# backbone_name = f"dinov2_{backbone_arch}"
+
+# backbone_model = torch.hub.load(repo_or_dir="facebookresearch/dinov2", model=backbone_name)
+# backbone_model.eval()
+# backbone_model.cuda()
+
+import urllib
+
+import mmcv
+from mmcv.runner import load_checkpoint
+
+
+def load_config_from_url(url: str) -> str:
+    with urllib.request.urlopen(url) as f:
+        return f.read().decode()
+
+
+# HEAD_SCALE_COUNT = 3 # more scales: slower but better results, in (1,2,3,4,5)
+# HEAD_DATASET = "voc2012" # in ("ade20k", "voc2012")
+# HEAD_TYPE = "ms" # in ("ms, "linear")
+
+
+DINOV2_BASE_URL = "https://dl.fbaipublicfiles.com/dinov2"
+# head_config_url = f"{DINOV2_BASE_URL}/{backbone_name}/{backbone_name}_{HEAD_DATASET}_{HEAD_TYPE}_config.py"
+# head_checkpoint_url = f"{DINOV2_BASE_URL}/{backbone_name}/{backbone_name}_{HEAD_DATASET}_{HEAD_TYPE}_head.pth"
+
+# cfg_str = load_config_from_url(head_config_url)
+# cfg = mmcv.Config.fromstring(cfg_str, file_format=".py")
+# if HEAD_TYPE == "ms":
+#     cfg.data.test.pipeline[1]["img_ratios"] = cfg.data.test.pipeline[1]["img_ratios"][:HEAD_SCALE_COUNT]
+#     print("scales:", cfg.data.test.pipeline[1]["img_ratios"])
+
+# model = create_segmenter(cfg, backbone_model=backbone_model)
+# load_checkpoint(model, head_checkpoint_url, map_location="cpu")
+# model.cuda()
+# model.eval()
+
+import urllib
+
+from PIL import Image
+
+
+def load_image_from_url(url: str) -> Image:
+    with urllib.request.urlopen(url) as f:
+        return Image.open(f).convert("RGB")
+
+
+EXAMPLE_IMAGE_URL = "https://dl.fbaipublicfiles.com/dinov2/images/example.jpg"
+
+
+# image = load_image_from_url(EXAMPLE_IMAGE_URL)
+image = Image.open("bridge_2.JPG").convert("RGB")
+
+image.show()
+
+import numpy as np
+
+import dinov2.eval.segmentation.utils.colormaps as colormaps
+
+
+DATASET_COLORMAPS = {
+    "ade20k": colormaps.ADE20K_COLORMAP,
+    "voc2012": colormaps.VOC2012_COLORMAP,
+}
+
+
+def render_segmentation(segmentation_logits, dataset):
+    colormap = DATASET_COLORMAPS[dataset]
+    colormap_array = np.array(colormap, dtype=np.uint8)
+    print(len(colormap))
+    segmentation_values = colormap_array[segmentation_logits + 1]
+    return Image.fromarray(segmentation_values)
+
+
+# array = np.array(image)[:, :, ::-1] # BGR
+# segmentation_logits = inference_segmentor(model, array)[0]
+# segmented_image = render_segmentation(segmentation_logits, HEAD_DATASET)
+# segmented_image.save("output.jpg")
+
+import dinov2.eval.segmentation_m2f.models.segmentors
+
+CONFIG_URL = f"{DINOV2_BASE_URL}/dinov2_vitg14/dinov2_vitg14_ade20k_m2f_config.py"
+CHECKPOINT_URL = f"{DINOV2_BASE_URL}/dinov2_vitg14/dinov2_vitg14_ade20k_m2f.pth"
+
+cfg_str = load_config_from_url(CONFIG_URL)
+cfg = mmcv.Config.fromstring(cfg_str, file_format=".py")
+
+model = init_segmentor(cfg)
+load_checkpoint(model, CHECKPOINT_URL, map_location="cpu")
+model.cuda()
+model.eval()
+
+array = np.array(image)[:, :, ::-1] # BGR
+segmentation_logits = inference_segmentor(model, array)[0]
+print(np.unique(segmentation_logits, return_counts=True))
+segmented_image = render_segmentation(segmentation_logits, "ade20k")
+segmented_image.save("output.jpg")

dinov2/__init__.py

@@ -0,0 +1,6 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+__version__ = "0.0.1"

dinov2/configs/__init__.py

@@ -0,0 +1,22 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import pathlib
+
+from omegaconf import OmegaConf
+
+
+def load_config(config_name: str):
+    config_filename = config_name + ".yaml"
+    return OmegaConf.load(pathlib.Path(__file__).parent.resolve() / config_filename)
+
+
+dinov2_default_config = load_config("ssl_default_config")
+
+
+def load_and_merge_config(config_name: str):
+    default_config = OmegaConf.create(dinov2_default_config)
+    loaded_config = load_config(config_name)
+    return OmegaConf.merge(default_config, loaded_config)

dinov2/configs/eval/vitb14_pretrain.yaml

@@ -0,0 +1,6 @@
+student:
+  arch: vit_base
+  patch_size: 14
+crops:
+  global_crops_size: 518  # this is to set up the position embeddings properly
+  local_crops_size: 98

dinov2/configs/eval/vitg14_pretrain.yaml

@@ -0,0 +1,7 @@
+student:
+  arch: vit_giant2
+  patch_size: 14
+  ffn_layer: swiglufused
+crops:
+  global_crops_size: 518  # this is to set up the position embeddings properly
+  local_crops_size: 98

dinov2/configs/eval/vitl14_pretrain.yaml

@@ -0,0 +1,6 @@
+student:
+  arch: vit_large
+  patch_size: 14
+crops:
+  global_crops_size: 518  # this is to set up the position embeddings properly
+  local_crops_size: 98

dinov2/configs/eval/vits14_pretrain.yaml

@@ -0,0 +1,6 @@
+student:
+  arch: vit_small
+  patch_size: 14
+crops:
+  global_crops_size: 518  # this is to set up the position embeddings properly
+  local_crops_size: 98

dinov2/configs/ssl_default_config.yaml

@@ -0,0 +1,115 @@
+MODEL:
+  WEIGHTS: ''
+compute_precision:
+  grad_scaler: true
+  teacher:
+    backbone:
+      sharding_strategy: SHARD_GRAD_OP
+      mixed_precision:
+        param_dtype: fp16
+        reduce_dtype: fp16
+        buffer_dtype: fp32
+    dino_head:
+      sharding_strategy: SHARD_GRAD_OP
+      mixed_precision:
+        param_dtype: fp16
+        reduce_dtype: fp16
+        buffer_dtype: fp32
+    ibot_head:
+      sharding_strategy: SHARD_GRAD_OP
+      mixed_precision:
+        param_dtype: fp16
+        reduce_dtype: fp16
+        buffer_dtype: fp32
+  student:
+    backbone:
+      sharding_strategy: SHARD_GRAD_OP
+      mixed_precision:
+        param_dtype: fp16
+        reduce_dtype: fp16
+        buffer_dtype: fp32
+    dino_head:
+      sharding_strategy: SHARD_GRAD_OP
+      mixed_precision:
+        param_dtype: fp16
+        reduce_dtype: fp32
+        buffer_dtype: fp32
+    ibot_head:
+      sharding_strategy: SHARD_GRAD_OP
+      mixed_precision:
+        param_dtype: fp16
+        reduce_dtype: fp32
+        buffer_dtype: fp32
+dino:
+  loss_weight: 1.0
+  head_n_prototypes: 65536
+  head_bottleneck_dim: 256
+  head_nlayers: 3
+  head_hidden_dim: 2048
+  koleo_loss_weight: 0.1
+ibot:
+  loss_weight: 1.0
+  mask_sample_probability: 0.5
+  mask_ratio_min_max:
+  - 0.1
+  - 0.5
+  separate_head: false
+  head_n_prototypes: 65536
+  head_bottleneck_dim: 256
+  head_nlayers: 3
+  head_hidden_dim: 2048
+train:
+  batch_size_per_gpu: 64
+  dataset_path: ImageNet:split=TRAIN
+  output_dir: .
+  saveckp_freq: 20
+  seed: 0
+  num_workers: 10
+  OFFICIAL_EPOCH_LENGTH: 1250
+  cache_dataset: true
+  centering: "centering" # or "sinkhorn_knopp"
+student:
+  arch: vit_large
+  patch_size: 16
+  drop_path_rate: 0.3
+  layerscale: 1.0e-05
+  drop_path_uniform: true
+  pretrained_weights: ''
+  ffn_layer: "mlp"
+  block_chunks: 0
+  qkv_bias: true
+  proj_bias: true
+  ffn_bias: true
+teacher:
+  momentum_teacher: 0.992
+  final_momentum_teacher: 1
+  warmup_teacher_temp: 0.04
+  teacher_temp: 0.07
+  warmup_teacher_temp_epochs: 30
+optim:
+  epochs: 100
+  weight_decay: 0.04
+  weight_decay_end: 0.4
+  base_lr: 0.004  # learning rate for a batch size of 1024
+  lr: 0.  # will be set after applying scaling rule
+  warmup_epochs: 10
+  min_lr: 1.0e-06
+  clip_grad: 3.0
+  freeze_last_layer_epochs: 1
+  scaling_rule: sqrt_wrt_1024
+  patch_embed_lr_mult: 0.2
+  layerwise_decay: 0.9
+  adamw_beta1: 0.9
+  adamw_beta2: 0.999
+crops:
+  global_crops_scale:
+  - 0.32
+  - 1.0
+  local_crops_number: 8
+  local_crops_scale:
+  - 0.05
+  - 0.32
+  global_crops_size: 224
+  local_crops_size: 96
+evaluation:
+  eval_period_iterations: 12500

dinov2/configs/train/vitg14.yaml

@@ -0,0 +1,26 @@
+dino:
+  head_n_prototypes: 131072
+  head_bottleneck_dim: 384
+ibot:
+  separate_head: true
+  head_n_prototypes: 131072
+train:
+  batch_size_per_gpu: 12
+  dataset_path: ImageNet22k
+  centering: sinkhorn_knopp
+student:
+  arch: vit_giant2
+  patch_size: 14
+  drop_path_rate: 0.4
+  ffn_layer: swiglufused
+  block_chunks: 4
+teacher:
+  momentum_teacher: 0.994
+optim:
+  epochs: 500
+  weight_decay_end: 0.2
+  base_lr: 2.0e-04  # learning rate for a batch size of 1024
+  warmup_epochs: 80
+  layerwise_decay: 1.0
+crops:
+  local_crops_size: 98

dinov2/configs/train/vitl14.yaml

@@ -0,0 +1,26 @@
+dino:
+  head_n_prototypes: 131072
+  head_bottleneck_dim: 384
+ibot:
+  separate_head: true
+  head_n_prototypes: 131072
+train:
+  batch_size_per_gpu: 32
+  dataset_path: ImageNet22k
+  centering: sinkhorn_knopp
+student:
+  arch: vit_large
+  patch_size: 14
+  drop_path_rate: 0.4
+  ffn_layer: swiglufused
+  block_chunks: 4
+teacher:
+  momentum_teacher: 0.994
+optim:
+  epochs: 500
+  weight_decay_end: 0.2
+  base_lr: 2.0e-04  # learning rate for a batch size of 1024
+  warmup_epochs: 80
+  layerwise_decay: 1.0
+crops:
+  local_crops_size: 98

dinov2/configs/train/vitl16_short.yaml

@@ -0,0 +1,6 @@
+# this corresponds to the default config
+train:
+  dataset_path: ImageNet:split=TRAIN
+  batch_size_per_gpu: 64
+student:
+  block_chunks: 4

dinov2/data/__init__.py

@@ -0,0 +1,10 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from .adapters import DatasetWithEnumeratedTargets
+from .loaders import make_data_loader, make_dataset, SamplerType
+from .collate import collate_data_and_cast
+from .masking import MaskingGenerator
+from .augmentations import DataAugmentationDINO

dinov2/data/adapters.py

@@ -0,0 +1,28 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from typing import Any, Tuple
+
+from torch.utils.data import Dataset
+
+
+class DatasetWithEnumeratedTargets(Dataset):
+    def __init__(self, dataset):
+        self._dataset = dataset
+
+    def get_image_data(self, index: int) -> bytes:
+        return self._dataset.get_image_data(index)
+
+    def get_target(self, index: int) -> Tuple[Any, int]:
+        target = self._dataset.get_target(index)
+        return (index, target)
+
+    def __getitem__(self, index: int) -> Tuple[Any, Tuple[Any, int]]:
+        image, target = self._dataset[index]
+        target = index if target is None else target
+        return image, (index, target)
+
+    def __len__(self) -> int:
+        return len(self._dataset)

dinov2/data/augmentations.py

@@ -0,0 +1,118 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import logging
+
+from torchvision import transforms
+
+from .transforms import (
+    GaussianBlur,
+    make_normalize_transform,
+)
+
+
+logger = logging.getLogger("dinov2")
+
+
+class DataAugmentationDINO(object):
+    def __init__(
+        self,
+        global_crops_scale,
+        local_crops_scale,
+        local_crops_number,
+        global_crops_size=224,
+        local_crops_size=96,
+    ):
+        self.global_crops_scale = global_crops_scale
+        self.local_crops_scale = local_crops_scale
+        self.local_crops_number = local_crops_number
+        self.global_crops_size = global_crops_size
+        self.local_crops_size = local_crops_size
+
+        logger.info("###################################")
+        logger.info("Using data augmentation parameters:")
+        logger.info(f"global_crops_scale: {global_crops_scale}")
+        logger.info(f"local_crops_scale: {local_crops_scale}")
+        logger.info(f"local_crops_number: {local_crops_number}")
+        logger.info(f"global_crops_size: {global_crops_size}")
+        logger.info(f"local_crops_size: {local_crops_size}")
+        logger.info("###################################")
+
+        # random resized crop and flip
+        self.geometric_augmentation_global = transforms.Compose(
+            [
+                transforms.RandomResizedCrop(
+                    global_crops_size, scale=global_crops_scale, interpolation=transforms.InterpolationMode.BICUBIC
+                ),
+                transforms.RandomHorizontalFlip(p=0.5),
+            ]
+        )
+
+        self.geometric_augmentation_local = transforms.Compose(
+            [
+                transforms.RandomResizedCrop(
+                    local_crops_size, scale=local_crops_scale, interpolation=transforms.InterpolationMode.BICUBIC
+                ),
+                transforms.RandomHorizontalFlip(p=0.5),
+            ]
+        )
+
+        # color distorsions / blurring
+        color_jittering = transforms.Compose(
+            [
+                transforms.RandomApply(
+                    [transforms.ColorJitter(brightness=0.4, contrast=0.4, saturation=0.2, hue=0.1)],
+                    p=0.8,
+                ),
+                transforms.RandomGrayscale(p=0.2),
+            ]
+        )
+
+        global_transfo1_extra = GaussianBlur(p=1.0)
+
+        global_transfo2_extra = transforms.Compose(
+            [
+                GaussianBlur(p=0.1),
+                transforms.RandomSolarize(threshold=128, p=0.2),
+            ]
+        )
+
+        local_transfo_extra = GaussianBlur(p=0.5)
+
+        # normalization
+        self.normalize = transforms.Compose(
+            [
+                transforms.ToTensor(),
+                make_normalize_transform(),
+            ]
+        )
+
+        self.global_transfo1 = transforms.Compose([color_jittering, global_transfo1_extra, self.normalize])
+        self.global_transfo2 = transforms.Compose([color_jittering, global_transfo2_extra, self.normalize])
+        self.local_transfo = transforms.Compose([color_jittering, local_transfo_extra, self.normalize])
+
+    def __call__(self, image):
+        output = {}
+
+        # global crops:
+        im1_base = self.geometric_augmentation_global(image)
+        global_crop_1 = self.global_transfo1(im1_base)
+
+        im2_base = self.geometric_augmentation_global(image)
+        global_crop_2 = self.global_transfo2(im2_base)
+
+        output["global_crops"] = [global_crop_1, global_crop_2]
+
+        # global crops for teacher:
+        output["global_crops_teacher"] = [global_crop_1, global_crop_2]
+
+        # local crops:
+        local_crops = [
+            self.local_transfo(self.geometric_augmentation_local(image)) for _ in range(self.local_crops_number)
+        ]
+        output["local_crops"] = local_crops
+        output["offsets"] = ()
+
+        return output

dinov2/data/collate.py

@@ -0,0 +1,49 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import torch
+import random
+
+
+def collate_data_and_cast(samples_list, mask_ratio_tuple, mask_probability, dtype, n_tokens=None, mask_generator=None):
+    # dtype = torch.half  # TODO: Remove
+
+    n_global_crops = len(samples_list[0][0]["global_crops"])
+    n_local_crops = len(samples_list[0][0]["local_crops"])
+
+    collated_global_crops = torch.stack([s[0]["global_crops"][i] for i in range(n_global_crops) for s in samples_list])
+
+    collated_local_crops = torch.stack([s[0]["local_crops"][i] for i in range(n_local_crops) for s in samples_list])
+
+    B = len(collated_global_crops)
+    N = n_tokens
+    n_samples_masked = int(B * mask_probability)
+    probs = torch.linspace(*mask_ratio_tuple, n_samples_masked + 1)
+    upperbound = 0
+    masks_list = []
+    for i in range(0, n_samples_masked):
+        prob_min = probs[i]
+        prob_max = probs[i + 1]
+        masks_list.append(torch.BoolTensor(mask_generator(int(N * random.uniform(prob_min, prob_max)))))
+        upperbound += int(N * prob_max)
+    for i in range(n_samples_masked, B):
+        masks_list.append(torch.BoolTensor(mask_generator(0)))
+
+    random.shuffle(masks_list)
+
+    collated_masks = torch.stack(masks_list).flatten(1)
+    mask_indices_list = collated_masks.flatten().nonzero().flatten()
+
+    masks_weight = (1 / collated_masks.sum(-1).clamp(min=1.0)).unsqueeze(-1).expand_as(collated_masks)[collated_masks]
+
+    return {
+        "collated_global_crops": collated_global_crops.to(dtype),
+        "collated_local_crops": collated_local_crops.to(dtype),
+        "collated_masks": collated_masks,
+        "mask_indices_list": mask_indices_list,
+        "masks_weight": masks_weight,
+        "upperbound": upperbound,
+        "n_masked_patches": torch.full((1,), fill_value=mask_indices_list.shape[0], dtype=torch.long),
+    }

dinov2/data/datasets/__init__.py

@@ -0,0 +1,7 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from .image_net import ImageNet
+from .image_net_22k import ImageNet22k

dinov2/data/datasets/decoders.py

@@ -0,0 +1,31 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from io import BytesIO
+from typing import Any
+
+from PIL import Image
+
+
+class Decoder:
+    def decode(self) -> Any:
+        raise NotImplementedError
+
+
+class ImageDataDecoder(Decoder):
+    def __init__(self, image_data: bytes) -> None:
+        self._image_data = image_data
+
+    def decode(self) -> Image:
+        f = BytesIO(self._image_data)
+        return Image.open(f).convert(mode="RGB")
+
+
+class TargetDecoder(Decoder):
+    def __init__(self, target: Any):
+        self._target = target
+
+    def decode(self) -> Any:
+        return self._target

dinov2/data/datasets/extended.py

@@ -0,0 +1,38 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from typing import Any, Tuple
+
+from torchvision.datasets import VisionDataset
+
+from .decoders import TargetDecoder, ImageDataDecoder
+
+
+class ExtendedVisionDataset(VisionDataset):
+    def __init__(self, *args, **kwargs) -> None:
+        super().__init__(*args, **kwargs)  # type: ignore
+
+    def get_image_data(self, index: int) -> bytes:
+        raise NotImplementedError
+
+    def get_target(self, index: int) -> Any:
+        raise NotImplementedError
+
+    def __getitem__(self, index: int) -> Tuple[Any, Any]:
+        try:
+            image_data = self.get_image_data(index)
+            image = ImageDataDecoder(image_data).decode()
+        except Exception as e:
+            raise RuntimeError(f"can not read image for sample {index}") from e
+        target = self.get_target(index)
+        target = TargetDecoder(target).decode()
+
+        if self.transforms is not None:
+            image, target = self.transforms(image, target)
+
+        return image, target
+
+    def __len__(self) -> int:
+        raise NotImplementedError

dinov2/data/datasets/image_net.py

@@ -0,0 +1,290 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import csv
+from enum import Enum
+import logging
+import os
+from typing import Callable, List, Optional, Tuple, Union
+
+import numpy as np
+
+from .extended import ExtendedVisionDataset
+
+
+logger = logging.getLogger("dinov2")
+_Target = int
+
+
+class _Split(Enum):
+    TRAIN = "train"
+    VAL = "val"
+    TEST = "test"  # NOTE: torchvision does not support the test split
+
+    @property
+    def length(self) -> int:
+        split_lengths = {
+            _Split.TRAIN: 1_281_167,
+            _Split.VAL: 50_000,
+            _Split.TEST: 100_000,
+        }
+        return split_lengths[self]
+
+    def get_dirname(self, class_id: Optional[str] = None) -> str:
+        return self.value if class_id is None else os.path.join(self.value, class_id)
+
+    def get_image_relpath(self, actual_index: int, class_id: Optional[str] = None) -> str:
+        dirname = self.get_dirname(class_id)
+        if self == _Split.TRAIN:
+            basename = f"{class_id}_{actual_index}"
+        else:  # self in (_Split.VAL, _Split.TEST):
+            basename = f"ILSVRC2012_{self.value}_{actual_index:08d}"
+        return os.path.join(dirname, basename + ".JPEG")
+
+    def parse_image_relpath(self, image_relpath: str) -> Tuple[str, int]:
+        assert self != _Split.TEST
+        dirname, filename = os.path.split(image_relpath)
+        class_id = os.path.split(dirname)[-1]
+        basename, _ = os.path.splitext(filename)
+        actual_index = int(basename.split("_")[-1])
+        return class_id, actual_index
+
+
+class ImageNet(ExtendedVisionDataset):
+    Target = Union[_Target]
+    Split = Union[_Split]
+
+    def __init__(
+        self,
+        *,
+        split: "ImageNet.Split",
+        root: str,
+        extra: str,
+        transforms: Optional[Callable] = None,
+        transform: Optional[Callable] = None,
+        target_transform: Optional[Callable] = None,
+    ) -> None:
+        super().__init__(root, transforms, transform, target_transform)
+        self._extra_root = extra
+        self._split = split
+
+        self._entries = None
+        self._class_ids = None
+        self._class_names = None
+
+    @property
+    def split(self) -> "ImageNet.Split":
+        return self._split
+
+    def _get_extra_full_path(self, extra_path: str) -> str:
+        return os.path.join(self._extra_root, extra_path)
+
+    def _load_extra(self, extra_path: str) -> np.ndarray:
+        extra_full_path = self._get_extra_full_path(extra_path)
+        return np.load(extra_full_path, mmap_mode="r")
+
+    def _save_extra(self, extra_array: np.ndarray, extra_path: str) -> None:
+        extra_full_path = self._get_extra_full_path(extra_path)
+        os.makedirs(self._extra_root, exist_ok=True)
+        np.save(extra_full_path, extra_array)
+
+    @property
+    def _entries_path(self) -> str:
+        return f"entries-{self._split.value.upper()}.npy"
+
+    @property
+    def _class_ids_path(self) -> str:
+        return f"class-ids-{self._split.value.upper()}.npy"
+
+    @property
+    def _class_names_path(self) -> str:
+        return f"class-names-{self._split.value.upper()}.npy"
+
+    def _get_entries(self) -> np.ndarray:
+        if self._entries is None:
+            self._entries = self._load_extra(self._entries_path)
+        assert self._entries is not None
+        return self._entries
+
+    def _get_class_ids(self) -> np.ndarray:
+        if self._split == _Split.TEST:
+            assert False, "Class IDs are not available in TEST split"
+        if self._class_ids is None:
+            self._class_ids = self._load_extra(self._class_ids_path)
+        assert self._class_ids is not None
+        return self._class_ids
+
+    def _get_class_names(self) -> np.ndarray:
+        if self._split == _Split.TEST:
+            assert False, "Class names are not available in TEST split"
+        if self._class_names is None:
+            self._class_names = self._load_extra(self._class_names_path)
+        assert self._class_names is not None
+        return self._class_names
+
+    def find_class_id(self, class_index: int) -> str:
+        class_ids = self._get_class_ids()
+        return str(class_ids[class_index])
+
+    def find_class_name(self, class_index: int) -> str:
+        class_names = self._get_class_names()
+        return str(class_names[class_index])
+
+    def get_image_data(self, index: int) -> bytes:
+        entries = self._get_entries()
+        actual_index = entries[index]["actual_index"]
+
+        class_id = self.get_class_id(index)
+
+        image_relpath = self.split.get_image_relpath(actual_index, class_id)
+        image_full_path = os.path.join(self.root, image_relpath)
+        with open(image_full_path, mode="rb") as f:
+            image_data = f.read()
+        return image_data
+
+    def get_target(self, index: int) -> Optional[Target]:
+        entries = self._get_entries()
+        class_index = entries[index]["class_index"]
+        return None if self.split == _Split.TEST else int(class_index)
+
+    def get_targets(self) -> Optional[np.ndarray]:
+        entries = self._get_entries()
+        return None if self.split == _Split.TEST else entries["class_index"]
+
+    def get_class_id(self, index: int) -> Optional[str]:
+        entries = self._get_entries()
+        class_id = entries[index]["class_id"]
+        return None if self.split == _Split.TEST else str(class_id)
+
+    def get_class_name(self, index: int) -> Optional[str]:
+        entries = self._get_entries()
+        class_name = entries[index]["class_name"]
+        return None if self.split == _Split.TEST else str(class_name)
+
+    def __len__(self) -> int:
+        entries = self._get_entries()
+        assert len(entries) == self.split.length
+        return len(entries)
+
+    def _load_labels(self, labels_path: str) -> List[Tuple[str, str]]:
+        labels_full_path = os.path.join(self.root, labels_path)
+        labels = []
+
+        try:
+            with open(labels_full_path, "r") as f:
+                reader = csv.reader(f)
+                for row in reader:
+                    class_id, class_name = row
+                    labels.append((class_id, class_name))
+        except OSError as e:
+            raise RuntimeError(f'can not read labels file "{labels_full_path}"') from e
+
+        return labels
+
+    def _dump_entries(self) -> None:
+        split = self.split
+        if split == ImageNet.Split.TEST:
+            dataset = None
+            sample_count = split.length
+            max_class_id_length, max_class_name_length = 0, 0
+        else:
+            labels_path = "labels.txt"
+            logger.info(f'loading labels from "{labels_path}"')
+            labels = self._load_labels(labels_path)
+
+            # NOTE: Using torchvision ImageFolder for consistency
+            from torchvision.datasets import ImageFolder
+
+            dataset_root = os.path.join(self.root, split.get_dirname())
+            dataset = ImageFolder(dataset_root)
+            sample_count = len(dataset)
+            max_class_id_length, max_class_name_length = -1, -1
+            for sample in dataset.samples:
+                _, class_index = sample
+                class_id, class_name = labels[class_index]
+                max_class_id_length = max(len(class_id), max_class_id_length)
+                max_class_name_length = max(len(class_name), max_class_name_length)
+
+        dtype = np.dtype(
+            [
+                ("actual_index", "<u4"),
+                ("class_index", "<u4"),
+                ("class_id", f"U{max_class_id_length}"),
+                ("class_name", f"U{max_class_name_length}"),
+            ]
+        )
+        entries_array = np.empty(sample_count, dtype=dtype)
+
+        if split == ImageNet.Split.TEST:
+            old_percent = -1
+            for index in range(sample_count):
+                percent = 100 * (index + 1) // sample_count
+                if percent > old_percent:
+                    logger.info(f"creating entries: {percent}%")
+                    old_percent = percent
+
+                actual_index = index + 1
+                class_index = np.uint32(-1)
+                class_id, class_name = "", ""
+                entries_array[index] = (actual_index, class_index, class_id, class_name)
+        else:
+            class_names = {class_id: class_name for class_id, class_name in labels}
+
+            assert dataset
+            old_percent = -1
+            for index in range(sample_count):
+                percent = 100 * (index + 1) // sample_count
+                if percent > old_percent:
+                    logger.info(f"creating entries: {percent}%")
+                    old_percent = percent
+
+                image_full_path, class_index = dataset.samples[index]
+                image_relpath = os.path.relpath(image_full_path, self.root)
+                class_id, actual_index = split.parse_image_relpath(image_relpath)
+                class_name = class_names[class_id]
+                entries_array[index] = (actual_index, class_index, class_id, class_name)
+
+        logger.info(f'saving entries to "{self._entries_path}"')
+        self._save_extra(entries_array, self._entries_path)
+
+    def _dump_class_ids_and_names(self) -> None:
+        split = self.split
+        if split == ImageNet.Split.TEST:
+            return
+
+        entries_array = self._load_extra(self._entries_path)
+
+        max_class_id_length, max_class_name_length, max_class_index = -1, -1, -1
+        for entry in entries_array:
+            class_index, class_id, class_name = (
+                entry["class_index"],
+                entry["class_id"],
+                entry["class_name"],
+            )
+            max_class_index = max(int(class_index), max_class_index)
+            max_class_id_length = max(len(str(class_id)), max_class_id_length)
+            max_class_name_length = max(len(str(class_name)), max_class_name_length)
+
+        class_count = max_class_index + 1
+        class_ids_array = np.empty(class_count, dtype=f"U{max_class_id_length}")
+        class_names_array = np.empty(class_count, dtype=f"U{max_class_name_length}")
+        for entry in entries_array:
+            class_index, class_id, class_name = (
+                entry["class_index"],
+                entry["class_id"],
+                entry["class_name"],
+            )
+            class_ids_array[class_index] = class_id
+            class_names_array[class_index] = class_name
+
+        logger.info(f'saving class IDs to "{self._class_ids_path}"')
+        self._save_extra(class_ids_array, self._class_ids_path)
+
+        logger.info(f'saving class names to "{self._class_names_path}"')
+        self._save_extra(class_names_array, self._class_names_path)
+
+    def dump_extra(self) -> None:
+        self._dump_entries()
+        self._dump_class_ids_and_names()

dinov2/data/datasets/image_net_22k.py

@@ -0,0 +1,302 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from dataclasses import dataclass
+from enum import Enum
+from functools import lru_cache
+from gzip import GzipFile
+from io import BytesIO
+from mmap import ACCESS_READ, mmap
+import os
+from typing import Any, Callable, List, Optional, Set, Tuple
+import warnings
+
+import numpy as np
+
+from .extended import ExtendedVisionDataset
+
+
+_Labels = int
+
+_DEFAULT_MMAP_CACHE_SIZE = 16  # Warning: This can exhaust file descriptors
+
+
+@dataclass
+class _ClassEntry:
+    block_offset: int
+    maybe_filename: Optional[str] = None
+
+
+@dataclass
+class _Entry:
+    class_index: int  # noqa: E701
+    start_offset: int
+    end_offset: int
+    filename: str
+
+
+class _Split(Enum):
+    TRAIN = "train"
+    VAL = "val"
+
+    @property
+    def length(self) -> int:
+        return {
+            _Split.TRAIN: 11_797_647,
+            _Split.VAL: 561_050,
+        }[self]
+
+    def entries_path(self):
+        return f"imagenet21kp_{self.value}.txt"
+
+
+def _get_tarball_path(class_id: str) -> str:
+    return f"{class_id}.tar"
+
+
+def _make_mmap_tarball(tarballs_root: str, mmap_cache_size: int):
+    @lru_cache(maxsize=mmap_cache_size)
+    def _mmap_tarball(class_id: str) -> mmap:
+        tarball_path = _get_tarball_path(class_id)
+        tarball_full_path = os.path.join(tarballs_root, tarball_path)
+        with open(tarball_full_path) as f:
+            return mmap(fileno=f.fileno(), length=0, access=ACCESS_READ)
+
+    return _mmap_tarball
+
+
+class ImageNet22k(ExtendedVisionDataset):
+    _GZIPPED_INDICES: Set[int] = {
+        841_545,
+        1_304_131,
+        2_437_921,
+        2_672_079,
+        2_795_676,
+        2_969_786,
+        6_902_965,
+        6_903_550,
+        6_903_628,
+        7_432_557,
+        7_432_589,
+        7_813_809,
+        8_329_633,
+        10_296_990,
+        10_417_652,
+        10_492_265,
+        10_598_078,
+        10_782_398,
+        10_902_612,
+        11_203_736,
+        11_342_890,
+        11_397_596,
+        11_589_762,
+        11_705_103,
+        12_936_875,
+        13_289_782,
+    }
+    Labels = _Labels
+
+    def __init__(
+        self,
+        *,
+        root: str,
+        extra: str,
+        transforms: Optional[Callable] = None,
+        transform: Optional[Callable] = None,
+        target_transform: Optional[Callable] = None,
+        mmap_cache_size: int = _DEFAULT_MMAP_CACHE_SIZE,
+    ) -> None:
+        super().__init__(root, transforms, transform, target_transform)
+        self._extra_root = extra
+
+        entries_path = self._get_entries_path(root)
+        self._entries = self._load_extra(entries_path)
+
+        class_ids_path = self._get_class_ids_path(root)
+        self._class_ids = self._load_extra(class_ids_path)
+
+        self._gzipped_indices = ImageNet22k._GZIPPED_INDICES
+        self._mmap_tarball = _make_mmap_tarball(self._tarballs_root, mmap_cache_size)
+
+    def _get_entries_path(self, root: Optional[str] = None) -> str:
+        return "entries.npy"
+
+    def _get_class_ids_path(self, root: Optional[str] = None) -> str:
+        return "class-ids.npy"
+
+    def _find_class_ids(self, path: str) -> List[str]:
+        class_ids = []
+
+        with os.scandir(path) as entries:
+            for entry in entries:
+                root, ext = os.path.splitext(entry.name)
+                if ext != ".tar":
+                    continue
+                class_ids.append(root)
+
+        return sorted(class_ids)
+
+    def _load_entries_class_ids(self, root: Optional[str] = None) -> Tuple[List[_Entry], List[str]]:
+        root = self.get_root(root)
+        entries: List[_Entry] = []
+        class_ids = self._find_class_ids(root)
+
+        for class_index, class_id in enumerate(class_ids):
+            path = os.path.join(root, "blocks", f"{class_id}.log")
+            class_entries = []
+
+            try:
+                with open(path) as f:
+                    for line in f:
+                        line = line.rstrip()
+                        block, filename = line.split(":")
+                        block_offset = int(block[6:])
+                        filename = filename[1:]
+
+                        maybe_filename = None
+                        if filename != "** Block of NULs **":
+                            maybe_filename = filename
+                            _, ext = os.path.splitext(filename)
+                            # assert ext == ".JPEG"
+
+                        class_entry = _ClassEntry(block_offset, maybe_filename)
+                        class_entries.append(class_entry)
+            except OSError as e:
+                raise RuntimeError(f'can not read blocks file "{path}"') from e
+
+            assert class_entries[-1].maybe_filename is None
+
+            for class_entry1, class_entry2 in zip(class_entries, class_entries[1:]):
+                assert class_entry1.block_offset <= class_entry2.block_offset
+                start_offset = 512 * class_entry1.block_offset
+                end_offset = 512 * class_entry2.block_offset
+                assert class_entry1.maybe_filename is not None
+                filename = class_entry1.maybe_filename
+                entry = _Entry(class_index, start_offset, end_offset, filename)
+                # Skip invalid image files (PIL throws UnidentifiedImageError)
+                if filename == "n06470073_47249.JPEG":
+                    continue
+                entries.append(entry)
+
+        return entries, class_ids
+
+    def _load_extra(self, extra_path: str) -> np.ndarray:
+        extra_root = self._extra_root
+        extra_full_path = os.path.join(extra_root, extra_path)
+        return np.load(extra_full_path, mmap_mode="r")
+
+    def _save_extra(self, extra_array: np.ndarray, extra_path: str) -> None:
+        extra_root = self._extra_root
+        extra_full_path = os.path.join(extra_root, extra_path)
+        os.makedirs(extra_root, exist_ok=True)
+        np.save(extra_full_path, extra_array)
+
+    @property
+    def _tarballs_root(self) -> str:
+        return self.root
+
+    def find_class_id(self, class_index: int) -> str:
+        return str(self._class_ids[class_index])
+
+    def get_image_data(self, index: int) -> bytes:
+        entry = self._entries[index]
+        class_id = entry["class_id"]
+        class_mmap = self._mmap_tarball(class_id)
+
+        start_offset, end_offset = entry["start_offset"], entry["end_offset"]
+        try:
+            mapped_data = class_mmap[start_offset:end_offset]
+            data = mapped_data[512:]  # Skip entry header block
+
+            if len(data) >= 2 and tuple(data[:2]) == (0x1F, 0x8B):
+                assert index in self._gzipped_indices, f"unexpected gzip header for sample {index}"
+                with GzipFile(fileobj=BytesIO(data)) as g:
+                    data = g.read()
+        except Exception as e:
+            raise RuntimeError(f"can not retrieve image data for sample {index} " f'from "{class_id}" tarball') from e
+
+        return data
+
+    def get_target(self, index: int) -> Any:
+        return int(self._entries[index]["class_index"])
+
+    def get_targets(self) -> np.ndarray:
+        return self._entries["class_index"]
+
+    def get_class_id(self, index: int) -> str:
+        return str(self._entries[index]["class_id"])
+
+    def get_class_ids(self) -> np.ndarray:
+        return self._entries["class_id"]
+
+    def __getitem__(self, index: int) -> Tuple[Any, Any]:
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore")
+            return super().__getitem__(index)
+
+    def __len__(self) -> int:
+        return len(self._entries)
+
+    def _dump_entries(self, *args, **kwargs) -> None:
+        entries, class_ids = self._load_entries_class_ids(*args, **kwargs)
+
+        max_class_id_length, max_filename_length, max_class_index = -1, -1, -1
+        for entry in entries:
+            class_id = class_ids[entry.class_index]
+            max_class_index = max(entry.class_index, max_class_index)
+            max_class_id_length = max(len(class_id), max_class_id_length)
+            max_filename_length = max(len(entry.filename), max_filename_length)
+
+        dtype = np.dtype(
+            [
+                ("class_index", "<u4"),
+                ("class_id", f"U{max_class_id_length}"),
+                ("start_offset", "<u4"),
+                ("end_offset", "<u4"),
+                ("filename", f"U{max_filename_length}"),
+            ]
+        )
+        sample_count = len(entries)
+        entries_array = np.empty(sample_count, dtype=dtype)
+        for i, entry in enumerate(entries):
+            class_index = entry.class_index
+            class_id = class_ids[class_index]
+            start_offset = entry.start_offset
+            end_offset = entry.end_offset
+            filename = entry.filename
+            entries_array[i] = (
+                class_index,
+                class_id,
+                start_offset,
+                end_offset,
+                filename,
+            )
+
+        entries_path = self._get_entries_path(*args, **kwargs)
+        self._save_extra(entries_array, entries_path)
+
+    def _dump_class_ids(self, *args, **kwargs) -> None:
+        entries_path = self._get_entries_path(*args, **kwargs)
+        entries_array = self._load_extra(entries_path)
+
+        max_class_id_length, max_class_index = -1, -1
+        for entry in entries_array:
+            class_index, class_id = entry["class_index"], entry["class_id"]
+            max_class_index = max(int(class_index), max_class_index)
+            max_class_id_length = max(len(str(class_id)), max_class_id_length)
+
+        class_ids_array = np.empty(max_class_index + 1, dtype=f"U{max_class_id_length}")
+        for entry in entries_array:
+            class_index, class_id = entry["class_index"], entry["class_id"]
+            class_ids_array[class_index] = class_id
+        class_ids_path = self._get_class_ids_path(*args, **kwargs)
+        self._save_extra(class_ids_array, class_ids_path)
+
+    def _dump_extra(self, *args, **kwargs) -> None:
+        self._dump_entries(*args, *kwargs)
+        self._dump_class_ids(*args, *kwargs)
+
+    def dump_extra(self, root: Optional[str] = None) -> None:
+        return self._dump_extra(root)

dinov2/data/loaders.py

@@ -0,0 +1,222 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import logging
+from enum import Enum
+from typing import Any, Callable, List, Optional, TypeVar
+
+import torch
+from torch.utils.data import Sampler
+
+from .datasets import ImageNet, ImageNet22k
+from .samplers import EpochSampler, InfiniteSampler, ShardedInfiniteSampler
+
+
+logger = logging.getLogger("dinov2")
+
+
+class SamplerType(Enum):
+    DISTRIBUTED = 0
+    EPOCH = 1
+    INFINITE = 2
+    SHARDED_INFINITE = 3
+    SHARDED_INFINITE_NEW = 4
+
+
+def _make_bool_str(b: bool) -> str:
+    return "yes" if b else "no"
+
+
+def _make_sample_transform(image_transform: Optional[Callable] = None, target_transform: Optional[Callable] = None):
+    def transform(sample):
+        image, target = sample
+        if image_transform is not None:
+            image = image_transform(image)
+        if target_transform is not None:
+            target = target_transform(target)
+        return image, target
+
+    return transform
+
+
+def _parse_dataset_str(dataset_str: str):
+    tokens = dataset_str.split(":")
+
+    name = tokens[0]
+    kwargs = {}
+
+    for token in tokens[1:]:
+        key, value = token.split("=")
+        assert key in ("root", "extra", "split")
+        kwargs[key] = value
+
+    if name == "ImageNet":
+        class_ = ImageNet
+        if "split" in kwargs:
+            kwargs["split"] = ImageNet.Split[kwargs["split"]]
+    elif name == "ImageNet22k":
+        class_ = ImageNet22k
+    else:
+        raise ValueError(f'Unsupported dataset "{name}"')
+
+    return class_, kwargs
+
+
+def make_dataset(
+    *,
+    dataset_str: str,
+    transform: Optional[Callable] = None,
+    target_transform: Optional[Callable] = None,
+):
+    """
+    Creates a dataset with the specified parameters.
+
+    Args:
+        dataset_str: A dataset string description (e.g. ImageNet:split=TRAIN).
+        transform: A transform to apply to images.
+        target_transform: A transform to apply to targets.
+
+    Returns:
+        The created dataset.
+    """
+    logger.info(f'using dataset: "{dataset_str}"')
+
+    class_, kwargs = _parse_dataset_str(dataset_str)
+    dataset = class_(transform=transform, target_transform=target_transform, **kwargs)
+
+    logger.info(f"# of dataset samples: {len(dataset):,d}")
+
+    # Aggregated datasets do not expose (yet) these attributes, so add them.
+    if not hasattr(dataset, "transform"):
+        setattr(dataset, "transform", transform)
+    if not hasattr(dataset, "target_transform"):
+        setattr(dataset, "target_transform", target_transform)
+
+    return dataset
+
+
+def _make_sampler(
+    *,
+    dataset,
+    type: Optional[SamplerType] = None,
+    shuffle: bool = False,
+    seed: int = 0,
+    size: int = -1,
+    advance: int = 0,
+) -> Optional[Sampler]:
+    sample_count = len(dataset)
+
+    if type == SamplerType.INFINITE:
+        logger.info("sampler: infinite")
+        if size > 0:
+            raise ValueError("sampler size > 0 is invalid")
+        return InfiniteSampler(
+            sample_count=sample_count,
+            shuffle=shuffle,
+            seed=seed,
+            advance=advance,
+        )
+    elif type in (SamplerType.SHARDED_INFINITE, SamplerType.SHARDED_INFINITE_NEW):
+        logger.info("sampler: sharded infinite")
+        if size > 0:
+            raise ValueError("sampler size > 0 is invalid")
+        # TODO: Remove support for old shuffling
+        use_new_shuffle_tensor_slice = type == SamplerType.SHARDED_INFINITE_NEW
+        return ShardedInfiniteSampler(
+            sample_count=sample_count,
+            shuffle=shuffle,
+            seed=seed,
+            advance=advance,
+            use_new_shuffle_tensor_slice=use_new_shuffle_tensor_slice,
+        )
+    elif type == SamplerType.EPOCH:
+        logger.info("sampler: epoch")
+        if advance > 0:
+            raise NotImplementedError("sampler advance > 0 is not supported")
+        size = size if size > 0 else sample_count
+        logger.info(f"# of samples / epoch: {size:,d}")
+        return EpochSampler(
+            size=size,
+            sample_count=sample_count,
+            shuffle=shuffle,
+            seed=seed,
+        )
+    elif type == SamplerType.DISTRIBUTED:
+        logger.info("sampler: distributed")
+        if size > 0:
+            raise ValueError("sampler size > 0 is invalid")
+        if advance > 0:
+            raise ValueError("sampler advance > 0 is invalid")
+        return torch.utils.data.DistributedSampler(
+            dataset=dataset,
+            shuffle=shuffle,
+            seed=seed,
+            drop_last=False,
+        )
+
+    logger.info("sampler: none")
+    return None
+
+
+T = TypeVar("T")
+
+
+def make_data_loader(
+    *,
+    dataset,
+    batch_size: int,
+    num_workers: int,
+    shuffle: bool = True,
+    seed: int = 0,
+    sampler_type: Optional[SamplerType] = SamplerType.INFINITE,
+    sampler_size: int = -1,
+    sampler_advance: int = 0,
+    drop_last: bool = True,
+    persistent_workers: bool = False,
+    collate_fn: Optional[Callable[[List[T]], Any]] = None,
+):
+    """
+    Creates a data loader with the specified parameters.
+
+    Args:
+        dataset: A dataset (third party, LaViDa or WebDataset).
+        batch_size: The size of batches to generate.
+        num_workers: The number of workers to use.
+        shuffle: Whether to shuffle samples.
+        seed: The random seed to use.
+        sampler_type: Which sampler to use: EPOCH, INFINITE, SHARDED_INFINITE, SHARDED_INFINITE_NEW, DISTRIBUTED or None.
+        sampler_size: The number of images per epoch (when applicable) or -1 for the entire dataset.
+        sampler_advance: How many samples to skip (when applicable).
+        drop_last: Whether the last non-full batch of data should be dropped.
+        persistent_workers: maintain the workers Dataset instances alive after a dataset has been consumed once.
+        collate_fn: Function that performs batch collation
+    """
+
+    sampler = _make_sampler(
+        dataset=dataset,
+        type=sampler_type,
+        shuffle=shuffle,
+        seed=seed,
+        size=sampler_size,
+        advance=sampler_advance,
+    )
+
+    logger.info("using PyTorch data loader")
+    data_loader = torch.utils.data.DataLoader(
+        dataset,
+        sampler=sampler,
+        batch_size=batch_size,
+        num_workers=num_workers,
+        pin_memory=True,
+        drop_last=drop_last,
+        persistent_workers=persistent_workers,
+        collate_fn=collate_fn,
+    )
+
+    try:
+        logger.info(f"# of batches: {len(data_loader):,d}")
+    except TypeError:  # data loader has no length
+        logger.info("infinite data loader")
+    return data_loader

dinov2/data/masking.py

@@ -0,0 +1,86 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import random
+import math
+import numpy as np
+
+
+class MaskingGenerator:
+    def __init__(
+        self,
+        input_size,
+        num_masking_patches=None,
+        min_num_patches=4,
+        max_num_patches=None,
+        min_aspect=0.3,
+        max_aspect=None,
+    ):
+        if not isinstance(input_size, tuple):
+            input_size = (input_size,) * 2
+        self.height, self.width = input_size
+
+        self.num_patches = self.height * self.width
+        self.num_masking_patches = num_masking_patches
+
+        self.min_num_patches = min_num_patches
+        self.max_num_patches = num_masking_patches if max_num_patches is None else max_num_patches
+
+        max_aspect = max_aspect or 1 / min_aspect
+        self.log_aspect_ratio = (math.log(min_aspect), math.log(max_aspect))
+
+    def __repr__(self):
+        repr_str = "Generator(%d, %d -> [%d ~ %d], max = %d, %.3f ~ %.3f)" % (
+            self.height,
+            self.width,
+            self.min_num_patches,
+            self.max_num_patches,
+            self.num_masking_patches,
+            self.log_aspect_ratio[0],
+            self.log_aspect_ratio[1],
+        )
+        return repr_str
+
+    def get_shape(self):
+        return self.height, self.width
+
+    def _mask(self, mask, max_mask_patches):
+        delta = 0
+        for _ in range(10):
+            target_area = random.uniform(self.min_num_patches, max_mask_patches)
+            aspect_ratio = math.exp(random.uniform(*self.log_aspect_ratio))
+            h = int(round(math.sqrt(target_area * aspect_ratio)))
+            w = int(round(math.sqrt(target_area / aspect_ratio)))
+            if w < self.width and h < self.height:
+                top = random.randint(0, self.height - h)
+                left = random.randint(0, self.width - w)
+
+                num_masked = mask[top : top + h, left : left + w].sum()
+                # Overlap
+                if 0 < h * w - num_masked <= max_mask_patches:
+                    for i in range(top, top + h):
+                        for j in range(left, left + w):
+                            if mask[i, j] == 0:
+                                mask[i, j] = 1
+                                delta += 1
+
+                if delta > 0:
+                    break
+        return delta
+
+    def __call__(self, num_masking_patches=0):
+        mask = np.zeros(shape=self.get_shape(), dtype=bool)
+        mask_count = 0
+        while mask_count < num_masking_patches:
+            max_mask_patches = num_masking_patches - mask_count
+            max_mask_patches = min(max_mask_patches, self.max_num_patches)
+
+            delta = self._mask(mask, max_mask_patches)
+            if delta == 0:
+                break
+            else:
+                mask_count += delta
+
+        return mask

dinov2/data/samplers.py

@@ -0,0 +1,229 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import itertools
+from typing import Any, Optional
+import warnings
+
+import numpy as np
+import torch
+from torch.utils.data.sampler import Sampler
+
+import dinov2.distributed as distributed
+
+
+class EpochSampler(Sampler):
+    def __init__(
+        self,
+        *,
+        size: int,
+        sample_count: int,
+        shuffle: bool = False,
+        seed: int = 0,
+        start: Optional[int] = None,
+        step: Optional[int] = None,
+    ):
+        self._size = size
+        self._sample_count = sample_count
+        self._shuffle = shuffle
+        self._seed = seed
+        self._start = distributed.get_global_rank() if start is None else start
+        self._step = distributed.get_global_size() if step is None else step
+        self._epoch = 0
+
+    def __iter__(self):
+        count = (self._size + self._sample_count - 1) // self._sample_count
+        tiled_indices = np.tile(np.arange(self._sample_count), count)
+        if self._shuffle:
+            seed = self._seed * self._epoch if self._seed != 0 else self._epoch
+            rng = np.random.default_rng(seed)
+            iterable = rng.choice(tiled_indices, self._size, replace=False)
+        else:
+            iterable = tiled_indices[: self._size]
+
+        yield from itertools.islice(iterable, self._start, None, self._step)
+
+    def __len__(self):
+        return (self._size - self._start + self._step - 1) // self._step
+
+    def set_epoch(self, epoch):
+        self._epoch = epoch
+
+
+def _get_numpy_dtype(size: int) -> Any:
+    return np.int32 if size <= 2**31 else np.int64
+
+
+def _get_torch_dtype(size: int) -> Any:
+    return torch.int32 if size <= 2**31 else torch.int64
+
+
+def _generate_randperm_indices(*, size: int, generator: torch.Generator):
+    """Generate the indices of a random permutation."""
+    dtype = _get_torch_dtype(size)
+    # This is actually matching PyTorch's CPU implementation, see: https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/TensorFactories.cpp#L900-L921
+    perm = torch.arange(size, dtype=dtype)
+    for i in range(size):
+        j = torch.randint(i, size, size=(1,), generator=generator).item()
+
+        # Always swap even if no-op
+        value = perm[j].item()
+        perm[j] = perm[i].item()
+        perm[i] = value
+        yield value
+
+
+class InfiniteSampler(Sampler):
+    def __init__(
+        self,
+        *,
+        sample_count: int,
+        shuffle: bool = False,
+        seed: int = 0,
+        start: Optional[int] = None,
+        step: Optional[int] = None,
+        advance: int = 0,
+    ):
+        self._sample_count = sample_count
+        self._seed = seed
+        self._shuffle = shuffle
+        self._start = distributed.get_global_rank() if start is None else start
+        self._step = distributed.get_global_size() if step is None else step
+        self._advance = advance
+
+    def __iter__(self):
+        if self._shuffle:
+            iterator = self._shuffled_iterator()
+        else:
+            iterator = self._iterator()
+
+        yield from itertools.islice(iterator, self._advance, None)
+
+    def _iterator(self):
+        assert not self._shuffle
+
+        while True:
+            iterable = range(self._sample_count)
+            yield from itertools.islice(iterable, self._start, None, self._step)
+
+    def _shuffled_iterator(self):
+        assert self._shuffle
+
+        # Instantiate a generator here (rather than in the ctor) to keep the class
+        # picklable (requirement of mp.spawn)
+        generator = torch.Generator().manual_seed(self._seed)
+
+        while True:
+            iterable = _generate_randperm_indices(size=self._sample_count, generator=generator)
+            yield from itertools.islice(iterable, self._start, None, self._step)
+
+
+# The following function is somewhat equivalent to _new_shuffle_tensor_slice below,
+# but avoids a full in-place random permutation generation.
+def _shuffle_tensor_slice(
+    *, tensor: torch.Tensor, start: int = 0, step: int = 1, generator: torch.Generator
+) -> np.ndarray:
+    stop = len(tensor)
+    count = stop // step
+    drop_count = stop - step * count
+    if drop_count:
+        warnings.warn(f"# of dropped samples: {drop_count}")
+
+    dtype = _get_numpy_dtype(stop)
+    result = np.empty(count, dtype=dtype)
+
+    for i in range(count):
+        j = torch.randint(0, i + 1, size=(1,), generator=generator).item() if i > 0 else 0
+
+        result[i] = result[j]
+        result[j] = tensor[start + i * step].item()
+
+    return result
+
+
+def _new_shuffle_tensor_slice(
+    *, tensor: torch.Tensor, start: int = 0, step: int = 1, generator: torch.Generator
+) -> np.ndarray:
+    stop = len(tensor)
+    count = stop // step
+    dtype = torch.int64  # Needed for using randperm result as indices
+    count = stop // step
+    drop_count = stop - step * count
+    if drop_count:
+        warnings.warn(f"# of dropped samples: {drop_count}")
+    indices = torch.randperm(count, dtype=dtype, generator=generator)
+    return tensor[start::step][indices].numpy()
+
+
+def _make_seed(seed: int, start: int, iter_count: int) -> int:
+    # NOTE: Tried a few variants (including iter_count << 32), this one worked best.
+    return seed + start + (iter_count << 24)
+
+
+class ShardedInfiniteSampler(Sampler):
+    def __init__(
+        self,
+        *,
+        sample_count: int,
+        shuffle: bool = False,
+        seed: int = 0,
+        start: Optional[int] = None,
+        step: Optional[int] = None,
+        advance: int = 0,
+        use_new_shuffle_tensor_slice: bool = False,
+    ):
+        self._sample_count = sample_count
+        self._seed = seed
+        self._shuffle = shuffle
+        self._start = distributed.get_global_rank() if start is None else start
+        self._step = distributed.get_global_size() if step is None else step
+        self._advance = advance
+        self._iter_count = 0
+        self._shuffle_tensor_slice_fn = (
+            _new_shuffle_tensor_slice if use_new_shuffle_tensor_slice else _shuffle_tensor_slice
+        )
+
+    def __iter__(self):
+        iter_count = self._advance // self._sample_count
+        if iter_count > 0:
+            self._advance -= iter_count * self._sample_count
+            self._iter_count += iter_count
+
+        if self._shuffle:
+            iterator = self._shuffled_iterator()
+        else:
+            iterator = self._iterator()
+
+        yield from itertools.islice(iterator, self._advance, None)
+
+    def _iterator(self):
+        assert not self._shuffle
+
+        while True:
+            iterable = range(self._sample_count)
+            yield from itertools.islice(iterable, self._start, None, self._step)
+
+    def _shuffled_iterator(self):
+        assert self._shuffle
+
+        # Instantiate a generator here (rather than in the ctor) to be keep the class
+        # picklable (requirement of mp.spawn)
+        generator = torch.Generator()
+
+        # Always shuffle everything first
+        generator.manual_seed(self._seed)
+        dtype = _get_torch_dtype(self._sample_count)
+        perm = torch.randperm(self._sample_count, dtype=dtype, generator=generator)
+
+        while True:
+            # Re-seed on each iteration to allow skipping whole permutations
+            seed = _make_seed(self._seed, self._start, self._iter_count)
+            generator.manual_seed(seed)
+
+            iterable = self._shuffle_tensor_slice_fn(
+                tensor=perm, start=self._start, step=self._step, generator=generator
+            )
+            yield from iterable
+            self._iter_count += 1

dinov2/data/transforms.py

@@ -0,0 +1,91 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from typing import Sequence
+
+import torch
+from torchvision import transforms
+
+
+class GaussianBlur(transforms.RandomApply):
+    """
+    Apply Gaussian Blur to the PIL image.
+    """
+
+    def __init__(self, *, p: float = 0.5, radius_min: float = 0.1, radius_max: float = 2.0):
+        # NOTE: torchvision is applying 1 - probability to return the original image
+        keep_p = 1 - p
+        transform = transforms.GaussianBlur(kernel_size=9, sigma=(radius_min, radius_max))
+        super().__init__(transforms=[transform], p=keep_p)
+
+
+class MaybeToTensor(transforms.ToTensor):
+    """
+    Convert a ``PIL Image`` or ``numpy.ndarray`` to tensor, or keep as is if already a tensor.
+    """
+
+    def __call__(self, pic):
+        """
+        Args:
+            pic (PIL Image, numpy.ndarray or torch.tensor): Image to be converted to tensor.
+        Returns:
+            Tensor: Converted image.
+        """
+        if isinstance(pic, torch.Tensor):
+            return pic
+        return super().__call__(pic)
+
+
+# Use timm's names
+IMAGENET_DEFAULT_MEAN = (0.485, 0.456, 0.406)
+IMAGENET_DEFAULT_STD = (0.229, 0.224, 0.225)
+
+
+def make_normalize_transform(
+    mean: Sequence[float] = IMAGENET_DEFAULT_MEAN,
+    std: Sequence[float] = IMAGENET_DEFAULT_STD,
+) -> transforms.Normalize:
+    return transforms.Normalize(mean=mean, std=std)
+
+
+# This roughly matches torchvision's preset for classification training:
+#   https://github.com/pytorch/vision/blob/main/references/classification/presets.py#L6-L44
+def make_classification_train_transform(
+    *,
+    crop_size: int = 224,
+    interpolation=transforms.InterpolationMode.BICUBIC,
+    hflip_prob: float = 0.5,
+    mean: Sequence[float] = IMAGENET_DEFAULT_MEAN,
+    std: Sequence[float] = IMAGENET_DEFAULT_STD,
+):
+    transforms_list = [transforms.RandomResizedCrop(crop_size, interpolation=interpolation)]
+    if hflip_prob > 0.0:
+        transforms_list.append(transforms.RandomHorizontalFlip(hflip_prob))
+    transforms_list.extend(
+        [
+            MaybeToTensor(),
+            make_normalize_transform(mean=mean, std=std),
+        ]
+    )
+    return transforms.Compose(transforms_list)
+
+
+# This matches (roughly) torchvision's preset for classification evaluation:
+#   https://github.com/pytorch/vision/blob/main/references/classification/presets.py#L47-L69
+def make_classification_eval_transform(
+    *,
+    resize_size: int = 256,
+    interpolation=transforms.InterpolationMode.BICUBIC,
+    crop_size: int = 224,
+    mean: Sequence[float] = IMAGENET_DEFAULT_MEAN,
+    std: Sequence[float] = IMAGENET_DEFAULT_STD,
+) -> transforms.Compose:
+    transforms_list = [
+        transforms.Resize(resize_size, interpolation=interpolation),
+        transforms.CenterCrop(crop_size),
+        MaybeToTensor(),
+        make_normalize_transform(mean=mean, std=std),
+    ]
+    return transforms.Compose(transforms_list)

dinov2/distributed/__init__.py

@@ -0,0 +1,270 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import os
+import random
+import re
+import socket
+from typing import Dict, List
+
+import torch
+import torch.distributed as dist
+
+_LOCAL_RANK = -1
+_LOCAL_WORLD_SIZE = -1
+
+
+def is_enabled() -> bool:
+    """
+    Returns:
+        True if distributed training is enabled
+    """
+    return dist.is_available() and dist.is_initialized()
+
+
+def get_global_size() -> int:
+    """
+    Returns:
+        The number of processes in the process group
+    """
+    return dist.get_world_size() if is_enabled() else 1
+
+
+def get_global_rank() -> int:
+    """
+    Returns:
+        The rank of the current process within the global process group.
+    """
+    return dist.get_rank() if is_enabled() else 0
+
+
+def get_local_rank() -> int:
+    """
+    Returns:
+        The rank of the current process within the local (per-machine) process group.
+    """
+    if not is_enabled():
+        return 0
+    assert 0 <= _LOCAL_RANK < _LOCAL_WORLD_SIZE
+    return _LOCAL_RANK
+
+
+def get_local_size() -> int:
+    """
+    Returns:
+        The size of the per-machine process group,
+        i.e. the number of processes per machine.
+    """
+    if not is_enabled():
+        return 1
+    assert 0 <= _LOCAL_RANK < _LOCAL_WORLD_SIZE
+    return _LOCAL_WORLD_SIZE
+
+
+def is_main_process() -> bool:
+    """
+    Returns:
+        True if the current process is the main one.
+    """
+    return get_global_rank() == 0
+
+
+def _restrict_print_to_main_process() -> None:
+    """
+    This function disables printing when not in the main process
+    """
+    import builtins as __builtin__
+
+    builtin_print = __builtin__.print
+
+    def print(*args, **kwargs):
+        force = kwargs.pop("force", False)
+        if is_main_process() or force:
+            builtin_print(*args, **kwargs)
+
+    __builtin__.print = print
+
+
+def _get_master_port(seed: int = 0) -> int:
+    MIN_MASTER_PORT, MAX_MASTER_PORT = (20_000, 60_000)
+
+    master_port_str = os.environ.get("MASTER_PORT")
+    if master_port_str is None:
+        rng = random.Random(seed)
+        return rng.randint(MIN_MASTER_PORT, MAX_MASTER_PORT)
+
+    return int(master_port_str)
+
+
+def _get_available_port() -> int:
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+        # A "" host address means INADDR_ANY i.e. binding to all interfaces.
+        # Note this is not compatible with IPv6.
+        s.bind(("", 0))
+        port = s.getsockname()[1]
+        return port
+
+
+_TORCH_DISTRIBUTED_ENV_VARS = (
+    "MASTER_ADDR",
+    "MASTER_PORT",
+    "RANK",
+    "WORLD_SIZE",
+    "LOCAL_RANK",
+    "LOCAL_WORLD_SIZE",
+)
+
+
+def _collect_env_vars() -> Dict[str, str]:
+    return {env_var: os.environ[env_var] for env_var in _TORCH_DISTRIBUTED_ENV_VARS if env_var in os.environ}
+
+
+def _is_slurm_job_process() -> bool:
+    return "SLURM_JOB_ID" in os.environ
+
+
+def _parse_slurm_node_list(s: str) -> List[str]:
+    nodes = []
+    # Extract "hostname", "hostname[1-2,3,4-5]," substrings
+    p = re.compile(r"(([^\[]+)(?:\[([^\]]+)\])?),?")
+    for m in p.finditer(s):
+        prefix, suffixes = s[m.start(2) : m.end(2)], s[m.start(3) : m.end(3)]
+        for suffix in suffixes.split(","):
+            span = suffix.split("-")
+            if len(span) == 1:
+                nodes.append(prefix + suffix)
+            else:
+                width = len(span[0])
+                start, end = int(span[0]), int(span[1]) + 1
+                nodes.extend([prefix + f"{i:0{width}}" for i in range(start, end)])
+    return nodes
+
+
+def _check_env_variable(key: str, new_value: str):
+    # Only check for difference with preset environment variables
+    if key in os.environ and os.environ[key] != new_value:
+        raise RuntimeError(f"Cannot export environment variables as {key} is already set")
+
+
+class _TorchDistributedEnvironment:
+    def __init__(self):
+        self.master_addr = "127.0.0.1"
+        self.master_port = 0
+        self.rank = -1
+        self.world_size = -1
+        self.local_rank = -1
+        self.local_world_size = -1
+
+        if _is_slurm_job_process():
+            return self._set_from_slurm_env()
+
+        env_vars = _collect_env_vars()
+        if not env_vars:
+            # Environment is not set
+            pass
+        elif len(env_vars) == len(_TORCH_DISTRIBUTED_ENV_VARS):
+            # Environment is fully set
+            return self._set_from_preset_env()
+        else:
+            # Environment is partially set
+            collected_env_vars = ", ".join(env_vars.keys())
+            raise RuntimeError(f"Partially set environment: {collected_env_vars}")
+
+        if torch.cuda.device_count() > 0:
+            return self._set_from_local()
+
+        raise RuntimeError("Can't initialize PyTorch distributed environment")
+
+    # Slurm job created with sbatch, submitit, etc...
+    def _set_from_slurm_env(self):
+        # logger.info("Initialization from Slurm environment")
+        job_id = int(os.environ["SLURM_JOB_ID"])
+        node_count = int(os.environ["SLURM_JOB_NUM_NODES"])
+        nodes = _parse_slurm_node_list(os.environ["SLURM_JOB_NODELIST"])
+        assert len(nodes) == node_count
+
+        self.master_addr = nodes[0]
+        self.master_port = _get_master_port(seed=job_id)
+        self.rank = int(os.environ["SLURM_PROCID"])
+        self.world_size = int(os.environ["SLURM_NTASKS"])
+        assert self.rank < self.world_size
+        self.local_rank = int(os.environ["SLURM_LOCALID"])
+        self.local_world_size = self.world_size // node_count
+        assert self.local_rank < self.local_world_size
+
+    # Single node job with preset environment (i.e. torchrun)
+    def _set_from_preset_env(self):
+        # logger.info("Initialization from preset environment")
+        self.master_addr = os.environ["MASTER_ADDR"]
+        self.master_port = os.environ["MASTER_PORT"]
+        self.rank = int(os.environ["RANK"])
+        self.world_size = int(os.environ["WORLD_SIZE"])
+        assert self.rank < self.world_size
+        self.local_rank = int(os.environ["LOCAL_RANK"])
+        self.local_world_size = int(os.environ["LOCAL_WORLD_SIZE"])
+        assert self.local_rank < self.local_world_size
+
+    # Single node and GPU job (i.e. local script run)
+    def _set_from_local(self):
+        # logger.info("Initialization from local")
+        self.master_addr = "127.0.0.1"
+        self.master_port = _get_available_port()
+        self.rank = 0
+        self.world_size = 1
+        self.local_rank = 0
+        self.local_world_size = 1
+
+    def export(self, *, overwrite: bool) -> "_TorchDistributedEnvironment":
+        # See the "Environment variable initialization" section from
+        # https://pytorch.org/docs/stable/distributed.html for the complete list of
+        # environment variables required for the env:// initialization method.
+        env_vars = {
+            "MASTER_ADDR": self.master_addr,
+            "MASTER_PORT": str(self.master_port),
+            "RANK": str(self.rank),
+            "WORLD_SIZE": str(self.world_size),
+            "LOCAL_RANK": str(self.local_rank),
+            "LOCAL_WORLD_SIZE": str(self.local_world_size),
+        }
+        if not overwrite:
+            for k, v in env_vars.items():
+                _check_env_variable(k, v)
+
+        os.environ.update(env_vars)
+        return self
+
+
+def enable(*, set_cuda_current_device: bool = True, overwrite: bool = False, allow_nccl_timeout: bool = False):
+    """Enable distributed mode
+
+    Args:
+        set_cuda_current_device: If True, call torch.cuda.set_device() to set the
+            current PyTorch CUDA device to the one matching the local rank.
+        overwrite: If True, overwrites already set variables. Else fails.
+    """
+
+    global _LOCAL_RANK, _LOCAL_WORLD_SIZE
+    if _LOCAL_RANK >= 0 or _LOCAL_WORLD_SIZE >= 0:
+        raise RuntimeError("Distributed mode has already been enabled")
+    torch_env = _TorchDistributedEnvironment()
+    torch_env.export(overwrite=overwrite)
+
+    if set_cuda_current_device:
+        torch.cuda.set_device(torch_env.local_rank)
+
+    if allow_nccl_timeout:
+        # This allows to use torch distributed timeout in a NCCL backend
+        key, value = "NCCL_ASYNC_ERROR_HANDLING", "1"
+        if not overwrite:
+            _check_env_variable(key, value)
+        os.environ[key] = value
+
+    dist.init_process_group(backend="nccl")
+    dist.barrier()
+
+    # Finalize setup
+    _LOCAL_RANK = torch_env.local_rank
+    _LOCAL_WORLD_SIZE = torch_env.local_world_size
+    _restrict_print_to_main_process()

dinov2/eval/__init__.py

@@ -0,0 +1,4 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.

dinov2/eval/depth/__init__.py

@@ -0,0 +1,4 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.

dinov2/eval/depth/models/__init__.py

@@ -0,0 +1,10 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from .backbones import *  # noqa: F403
+from .builder import BACKBONES, DEPTHER, HEADS, LOSSES, build_backbone, build_depther, build_head, build_loss
+from .decode_heads import *  # noqa: F403
+from .depther import *  # noqa: F403
+from .losses import *  # noqa: F403

dinov2/eval/depth/models/backbones/__init__.py

@@ -0,0 +1,6 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from .vision_transformer import DinoVisionTransformer

dinov2/eval/depth/models/backbones/vision_transformer.py

@@ -0,0 +1,16 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from mmcv.runner import BaseModule
+
+from ..builder import BACKBONES
+
+
+@BACKBONES.register_module()
+class DinoVisionTransformer(BaseModule):
+    """Vision Transformer."""
+
+    def __init__(self, *args, **kwargs):
+        super().__init__()

dinov2/eval/depth/models/builder.py

@@ -0,0 +1,49 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import warnings
+
+from mmcv.cnn import MODELS as MMCV_MODELS
+from mmcv.cnn.bricks.registry import ATTENTION as MMCV_ATTENTION
+from mmcv.utils import Registry
+
+MODELS = Registry("models", parent=MMCV_MODELS)
+ATTENTION = Registry("attention", parent=MMCV_ATTENTION)
+
+
+BACKBONES = MODELS
+NECKS = MODELS
+HEADS = MODELS
+LOSSES = MODELS
+DEPTHER = MODELS
+
+
+def build_backbone(cfg):
+    """Build backbone."""
+    return BACKBONES.build(cfg)
+
+
+def build_neck(cfg):
+    """Build neck."""
+    return NECKS.build(cfg)
+
+
+def build_head(cfg):
+    """Build head."""
+    return HEADS.build(cfg)
+
+
+def build_loss(cfg):
+    """Build loss."""
+    return LOSSES.build(cfg)
+
+
+def build_depther(cfg, train_cfg=None, test_cfg=None):
+    """Build depther."""
+    if train_cfg is not None or test_cfg is not None:
+        warnings.warn("train_cfg and test_cfg is deprecated, " "please specify them in model", UserWarning)
+    assert cfg.get("train_cfg") is None or train_cfg is None, "train_cfg specified in both outer field and model field "
+    assert cfg.get("test_cfg") is None or test_cfg is None, "test_cfg specified in both outer field and model field "
+    return DEPTHER.build(cfg, default_args=dict(train_cfg=train_cfg, test_cfg=test_cfg))

dinov2/eval/depth/models/decode_heads/__init__.py

@@ -0,0 +1,7 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+from .dpt_head import DPTHead
+from .linear_head import BNHead

dinov2/eval/depth/models/decode_heads/decode_head.py

@@ -0,0 +1,225 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import copy
+from abc import ABCMeta, abstractmethod
+
+import mmcv
+import numpy as np
+import torch
+import torch.nn as nn
+from mmcv.runner import BaseModule, auto_fp16, force_fp32
+
+from ...ops import resize
+from ..builder import build_loss
+
+
+class DepthBaseDecodeHead(BaseModule, metaclass=ABCMeta):
+    """Base class for BaseDecodeHead.
+
+    Args:
+        in_channels (List): Input channels.
+        channels (int): Channels after modules, before conv_depth.
+        conv_cfg (dict|None): Config of conv layers. Default: None.
+        act_cfg (dict): Config of activation layers.
+            Default: dict(type='ReLU')
+        loss_decode (dict): Config of decode loss.
+            Default: dict(type='SigLoss').
+        sampler (dict|None): The config of depth map sampler.
+            Default: None.
+        align_corners (bool): align_corners argument of F.interpolate.
+            Default: False.
+        min_depth (int): Min depth in dataset setting.
+            Default: 1e-3.
+        max_depth (int): Max depth in dataset setting.
+            Default: None.
+        norm_cfg (dict|None): Config of norm layers.
+            Default: None.
+        classify (bool): Whether predict depth in a cls.-reg. manner.
+            Default: False.
+        n_bins (int): The number of bins used in cls. step.
+            Default: 256.
+        bins_strategy (str): The discrete strategy used in cls. step.
+            Default: 'UD'.
+        norm_strategy (str): The norm strategy on cls. probability
+            distribution. Default: 'linear'
+        scale_up (str): Whether predict depth in a scale-up manner.
+            Default: False.
+    """
+
+    def __init__(
+        self,
+        in_channels,
+        channels=96,
+        conv_cfg=None,
+        act_cfg=dict(type="ReLU"),
+        loss_decode=dict(type="SigLoss", valid_mask=True, loss_weight=10),
+        sampler=None,
+        align_corners=False,
+        min_depth=1e-3,
+        max_depth=None,
+        norm_cfg=None,
+        classify=False,
+        n_bins=256,
+        bins_strategy="UD",
+        norm_strategy="linear",
+        scale_up=False,
+    ):
+        super(DepthBaseDecodeHead, self).__init__()
+
+        self.in_channels = in_channels
+        self.channels = channels
+        self.conv_cfg = conv_cfg
+        self.act_cfg = act_cfg
+        if isinstance(loss_decode, dict):
+            self.loss_decode = build_loss(loss_decode)
+        elif isinstance(loss_decode, (list, tuple)):
+            self.loss_decode = nn.ModuleList()
+            for loss in loss_decode:
+                self.loss_decode.append(build_loss(loss))
+        self.align_corners = align_corners
+        self.min_depth = min_depth
+        self.max_depth = max_depth
+        self.norm_cfg = norm_cfg
+        self.classify = classify
+        self.n_bins = n_bins
+        self.scale_up = scale_up
+
+        if self.classify:
+            assert bins_strategy in ["UD", "SID"], "Support bins_strategy: UD, SID"
+            assert norm_strategy in ["linear", "softmax", "sigmoid"], "Support norm_strategy: linear, softmax, sigmoid"
+
+            self.bins_strategy = bins_strategy
+            self.norm_strategy = norm_strategy
+            self.softmax = nn.Softmax(dim=1)
+            self.conv_depth = nn.Conv2d(channels, n_bins, kernel_size=3, padding=1, stride=1)
+        else:
+            self.conv_depth = nn.Conv2d(channels, 1, kernel_size=3, padding=1, stride=1)
+
+        self.fp16_enabled = False
+        self.relu = nn.ReLU()
+        self.sigmoid = nn.Sigmoid()
+
+    def extra_repr(self):
+        """Extra repr."""
+        s = f"align_corners={self.align_corners}"
+        return s
+
+    @auto_fp16()
+    @abstractmethod
+    def forward(self, inputs, img_metas):
+        """Placeholder of forward function."""
+        pass
+
+    def forward_train(self, img, inputs, img_metas, depth_gt, train_cfg):
+        """Forward function for training.
+        Args:
+            inputs (list[Tensor]): List of multi-level img features.
+            img_metas (list[dict]): List of image info dict where each dict
+                has: 'img_shape', 'scale_factor', 'flip', and may also contain
+                'filename', 'ori_shape', 'pad_shape', and 'img_norm_cfg'.
+                For details on the values of these keys see
+                `depth/datasets/pipelines/formatting.py:Collect`.
+            depth_gt (Tensor): GT depth
+            train_cfg (dict): The training config.
+
+        Returns:
+            dict[str, Tensor]: a dictionary of loss components
+        """
+        depth_pred = self.forward(inputs, img_metas)
+        losses = self.losses(depth_pred, depth_gt)
+
+        log_imgs = self.log_images(img[0], depth_pred[0], depth_gt[0], img_metas[0])
+        losses.update(**log_imgs)
+
+        return losses
+
+    def forward_test(self, inputs, img_metas, test_cfg):
+        """Forward function for testing.
+        Args:
+            inputs (list[Tensor]): List of multi-level img features.
+            img_metas (list[dict]): List of image info dict where each dict
+                has: 'img_shape', 'scale_factor', 'flip', and may also contain
+                'filename', 'ori_shape', 'pad_shape', and 'img_norm_cfg'.
+                For details on the values of these keys see
+                `depth/datasets/pipelines/formatting.py:Collect`.
+            test_cfg (dict): The testing config.
+
+        Returns:
+            Tensor: Output depth map.
+        """
+        return self.forward(inputs, img_metas)
+
+    def depth_pred(self, feat):
+        """Prediction each pixel."""
+        if self.classify:
+            logit = self.conv_depth(feat)
+
+            if self.bins_strategy == "UD":
+                bins = torch.linspace(self.min_depth, self.max_depth, self.n_bins, device=feat.device)
+            elif self.bins_strategy == "SID":
+                bins = torch.logspace(self.min_depth, self.max_depth, self.n_bins, device=feat.device)
+
+            # following Adabins, default linear
+            if self.norm_strategy == "linear":
+                logit = torch.relu(logit)
+                eps = 0.1
+                logit = logit + eps
+                logit = logit / logit.sum(dim=1, keepdim=True)
+            elif self.norm_strategy == "softmax":
+                logit = torch.softmax(logit, dim=1)
+            elif self.norm_strategy == "sigmoid":
+                logit = torch.sigmoid(logit)
+                logit = logit / logit.sum(dim=1, keepdim=True)
+
+            output = torch.einsum("ikmn,k->imn", [logit, bins]).unsqueeze(dim=1)
+
+        else:
+            if self.scale_up:
+                output = self.sigmoid(self.conv_depth(feat)) * self.max_depth
+            else:
+                output = self.relu(self.conv_depth(feat)) + self.min_depth
+        return output
+
+    @force_fp32(apply_to=("depth_pred",))
+    def losses(self, depth_pred, depth_gt):
+        """Compute depth loss."""
+        loss = dict()
+        depth_pred = resize(
+            input=depth_pred, size=depth_gt.shape[2:], mode="bilinear", align_corners=self.align_corners, warning=False
+        )
+        if not isinstance(self.loss_decode, nn.ModuleList):
+            losses_decode = [self.loss_decode]
+        else:
+            losses_decode = self.loss_decode
+        for loss_decode in losses_decode:
+            if loss_decode.loss_name not in loss:
+                loss[loss_decode.loss_name] = loss_decode(depth_pred, depth_gt)
+            else:
+                loss[loss_decode.loss_name] += loss_decode(depth_pred, depth_gt)
+        return loss
+
+    def log_images(self, img_path, depth_pred, depth_gt, img_meta):
+        show_img = copy.deepcopy(img_path.detach().cpu().permute(1, 2, 0))
+        show_img = show_img.numpy().astype(np.float32)
+        show_img = mmcv.imdenormalize(
+            show_img,
+            img_meta["img_norm_cfg"]["mean"],
+            img_meta["img_norm_cfg"]["std"],
+            img_meta["img_norm_cfg"]["to_rgb"],
+        )
+        show_img = np.clip(show_img, 0, 255)
+        show_img = show_img.astype(np.uint8)
+        show_img = show_img[:, :, ::-1]
+        show_img = show_img.transpose(0, 2, 1)
+        show_img = show_img.transpose(1, 0, 2)
+
+        depth_pred = depth_pred / torch.max(depth_pred)
+        depth_gt = depth_gt / torch.max(depth_gt)
+
+        depth_pred_color = copy.deepcopy(depth_pred.detach().cpu())
+        depth_gt_color = copy.deepcopy(depth_gt.detach().cpu())
+
+        return {"img_rgb": show_img, "img_depth_pred": depth_pred_color, "img_depth_gt": depth_gt_color}

dinov2/eval/depth/models/decode_heads/dpt_head.py

@@ -0,0 +1,270 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import math
+
+import torch
+import torch.nn as nn
+from mmcv.cnn import ConvModule, Linear, build_activation_layer
+from mmcv.runner import BaseModule
+
+from ...ops import resize
+from ..builder import HEADS
+from .decode_head import DepthBaseDecodeHead
+
+
+class Interpolate(nn.Module):
+    def __init__(self, scale_factor, mode, align_corners=False):
+        super(Interpolate, self).__init__()
+        self.interp = nn.functional.interpolate
+        self.scale_factor = scale_factor
+        self.mode = mode
+        self.align_corners = align_corners
+
+    def forward(self, x):
+        x = self.interp(x, scale_factor=self.scale_factor, mode=self.mode, align_corners=self.align_corners)
+        return x
+
+
+class HeadDepth(nn.Module):
+    def __init__(self, features):
+        super(HeadDepth, self).__init__()
+        self.head = nn.Sequential(
+            nn.Conv2d(features, features // 2, kernel_size=3, stride=1, padding=1),
+            Interpolate(scale_factor=2, mode="bilinear", align_corners=True),
+            nn.Conv2d(features // 2, 32, kernel_size=3, stride=1, padding=1),
+            nn.ReLU(),
+            nn.Conv2d(32, 1, kernel_size=1, stride=1, padding=0),
+        )
+
+    def forward(self, x):
+        x = self.head(x)
+        return x
+
+
+class ReassembleBlocks(BaseModule):
+    """ViTPostProcessBlock, process cls_token in ViT backbone output and
+    rearrange the feature vector to feature map.
+    Args:
+        in_channels (int): ViT feature channels. Default: 768.
+        out_channels (List): output channels of each stage.
+            Default: [96, 192, 384, 768].
+        readout_type (str): Type of readout operation. Default: 'ignore'.
+        patch_size (int): The patch size. Default: 16.
+        init_cfg (dict, optional): Initialization config dict. Default: None.
+    """
+
+    def __init__(
+        self, in_channels=768, out_channels=[96, 192, 384, 768], readout_type="ignore", patch_size=16, init_cfg=None
+    ):
+        super(ReassembleBlocks, self).__init__(init_cfg)
+
+        assert readout_type in ["ignore", "add", "project"]
+        self.readout_type = readout_type
+        self.patch_size = patch_size
+
+        self.projects = nn.ModuleList(
+            [
+                ConvModule(
+                    in_channels=in_channels,
+                    out_channels=out_channel,
+                    kernel_size=1,
+                    act_cfg=None,
+                )
+                for out_channel in out_channels
+            ]
+        )
+
+        self.resize_layers = nn.ModuleList(
+            [
+                nn.ConvTranspose2d(
+                    in_channels=out_channels[0], out_channels=out_channels[0], kernel_size=4, stride=4, padding=0
+                ),
+                nn.ConvTranspose2d(
+                    in_channels=out_channels[1], out_channels=out_channels[1], kernel_size=2, stride=2, padding=0
+                ),
+                nn.Identity(),
+                nn.Conv2d(
+                    in_channels=out_channels[3], out_channels=out_channels[3], kernel_size=3, stride=2, padding=1
+                ),
+            ]
+        )
+        if self.readout_type == "project":
+            self.readout_projects = nn.ModuleList()
+            for _ in range(len(self.projects)):
+                self.readout_projects.append(
+                    nn.Sequential(Linear(2 * in_channels, in_channels), build_activation_layer(dict(type="GELU")))
+                )
+
+    def forward(self, inputs):
+        assert isinstance(inputs, list)
+        out = []
+        for i, x in enumerate(inputs):
+            assert len(x) == 2
+            x, cls_token = x[0], x[1]
+            feature_shape = x.shape
+            if self.readout_type == "project":
+                x = x.flatten(2).permute((0, 2, 1))
+                readout = cls_token.unsqueeze(1).expand_as(x)
+                x = self.readout_projects[i](torch.cat((x, readout), -1))
+                x = x.permute(0, 2, 1).reshape(feature_shape)
+            elif self.readout_type == "add":
+                x = x.flatten(2) + cls_token.unsqueeze(-1)
+                x = x.reshape(feature_shape)
+            else:
+                pass
+            x = self.projects[i](x)
+            x = self.resize_layers[i](x)
+            out.append(x)
+        return out
+
+
+class PreActResidualConvUnit(BaseModule):
+    """ResidualConvUnit, pre-activate residual unit.
+    Args:
+        in_channels (int): number of channels in the input feature map.
+        act_cfg (dict): dictionary to construct and config activation layer.
+        norm_cfg (dict): dictionary to construct and config norm layer.
+        stride (int): stride of the first block. Default: 1
+        dilation (int): dilation rate for convs layers. Default: 1.
+        init_cfg (dict, optional): Initialization config dict. Default: None.
+    """
+
+    def __init__(self, in_channels, act_cfg, norm_cfg, stride=1, dilation=1, init_cfg=None):
+        super(PreActResidualConvUnit, self).__init__(init_cfg)
+
+        self.conv1 = ConvModule(
+            in_channels,
+            in_channels,
+            3,
+            stride=stride,
+            padding=dilation,
+            dilation=dilation,
+            norm_cfg=norm_cfg,
+            act_cfg=act_cfg,
+            bias=False,
+            order=("act", "conv", "norm"),
+        )
+
+        self.conv2 = ConvModule(
+            in_channels,
+            in_channels,
+            3,
+            padding=1,
+            norm_cfg=norm_cfg,
+            act_cfg=act_cfg,
+            bias=False,
+            order=("act", "conv", "norm"),
+        )
+
+    def forward(self, inputs):
+        inputs_ = inputs.clone()
+        x = self.conv1(inputs)
+        x = self.conv2(x)
+        return x + inputs_
+
+
+class FeatureFusionBlock(BaseModule):
+    """FeatureFusionBlock, merge feature map from different stages.
+    Args:
+        in_channels (int): Input channels.
+        act_cfg (dict): The activation config for ResidualConvUnit.
+        norm_cfg (dict): Config dict for normalization layer.
+        expand (bool): Whether expand the channels in post process block.
+            Default: False.
+        align_corners (bool): align_corner setting for bilinear upsample.
+            Default: True.
+        init_cfg (dict, optional): Initialization config dict. Default: None.
+    """
+
+    def __init__(self, in_channels, act_cfg, norm_cfg, expand=False, align_corners=True, init_cfg=None):
+        super(FeatureFusionBlock, self).__init__(init_cfg)
+
+        self.in_channels = in_channels
+        self.expand = expand
+        self.align_corners = align_corners
+
+        self.out_channels = in_channels
+        if self.expand:
+            self.out_channels = in_channels // 2
+
+        self.project = ConvModule(self.in_channels, self.out_channels, kernel_size=1, act_cfg=None, bias=True)
+
+        self.res_conv_unit1 = PreActResidualConvUnit(in_channels=self.in_channels, act_cfg=act_cfg, norm_cfg=norm_cfg)
+        self.res_conv_unit2 = PreActResidualConvUnit(in_channels=self.in_channels, act_cfg=act_cfg, norm_cfg=norm_cfg)
+
+    def forward(self, *inputs):
+        x = inputs[0]
+        if len(inputs) == 2:
+            if x.shape != inputs[1].shape:
+                res = resize(inputs[1], size=(x.shape[2], x.shape[3]), mode="bilinear", align_corners=False)
+            else:
+                res = inputs[1]
+            x = x + self.res_conv_unit1(res)
+        x = self.res_conv_unit2(x)
+        x = resize(x, scale_factor=2, mode="bilinear", align_corners=self.align_corners)
+        x = self.project(x)
+        return x
+
+
+@HEADS.register_module()
+class DPTHead(DepthBaseDecodeHead):
+    """Vision Transformers for Dense Prediction.
+    This head is implemented of `DPT <https://arxiv.org/abs/2103.13413>`_.
+    Args:
+        embed_dims (int): The embed dimension of the ViT backbone.
+            Default: 768.
+        post_process_channels (List): Out channels of post process conv
+            layers. Default: [96, 192, 384, 768].
+        readout_type (str): Type of readout operation. Default: 'ignore'.
+        patch_size (int): The patch size. Default: 16.
+        expand_channels (bool): Whether expand the channels in post process
+            block. Default: False.
+    """
+
+    def __init__(
+        self,
+        embed_dims=768,
+        post_process_channels=[96, 192, 384, 768],
+        readout_type="ignore",
+        patch_size=16,
+        expand_channels=False,
+        **kwargs
+    ):
+        super(DPTHead, self).__init__(**kwargs)
+
+        self.in_channels = self.in_channels
+        self.expand_channels = expand_channels
+        self.reassemble_blocks = ReassembleBlocks(embed_dims, post_process_channels, readout_type, patch_size)
+
+        self.post_process_channels = [
+            channel * math.pow(2, i) if expand_channels else channel for i, channel in enumerate(post_process_channels)
+        ]
+        self.convs = nn.ModuleList()
+        for channel in self.post_process_channels:
+            self.convs.append(ConvModule(channel, self.channels, kernel_size=3, padding=1, act_cfg=None, bias=False))
+        self.fusion_blocks = nn.ModuleList()
+        for _ in range(len(self.convs)):
+            self.fusion_blocks.append(FeatureFusionBlock(self.channels, self.act_cfg, self.norm_cfg))
+        self.fusion_blocks[0].res_conv_unit1 = None
+        self.project = ConvModule(self.channels, self.channels, kernel_size=3, padding=1, norm_cfg=self.norm_cfg)
+        self.num_fusion_blocks = len(self.fusion_blocks)
+        self.num_reassemble_blocks = len(self.reassemble_blocks.resize_layers)
+        self.num_post_process_channels = len(self.post_process_channels)
+        assert self.num_fusion_blocks == self.num_reassemble_blocks
+        assert self.num_reassemble_blocks == self.num_post_process_channels
+        self.conv_depth = HeadDepth(self.channels)
+
+    def forward(self, inputs, img_metas):
+        assert len(inputs) == self.num_reassemble_blocks
+        x = [inp for inp in inputs]
+        x = self.reassemble_blocks(x)
+        x = [self.convs[i](feature) for i, feature in enumerate(x)]
+        out = self.fusion_blocks[0](x[-1])
+        for i in range(1, len(self.fusion_blocks)):
+            out = self.fusion_blocks[i](out, x[-(i + 1)])
+        out = self.project(out)
+        out = self.depth_pred(out)
+        return out

dinov2/eval/depth/models/decode_heads/linear_head.py

@@ -0,0 +1,89 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This source code is licensed under the Apache License, Version 2.0
+# found in the LICENSE file in the root directory of this source tree.
+
+import torch
+import torch.nn as nn
+
+from ...ops import resize
+from ..builder import HEADS
+from .decode_head import DepthBaseDecodeHead
+
+
+@HEADS.register_module()
+class BNHead(DepthBaseDecodeHead):
+    """Just a batchnorm."""
+
+    def __init__(self, input_transform="resize_concat", in_index=(0, 1, 2, 3), upsample=1, **kwargs):
+        super().__init__(**kwargs)
+        self.input_transform = input_transform
+        self.in_index = in_index
+        self.upsample = upsample
+        # self.bn = nn.SyncBatchNorm(self.in_channels)
+        if self.classify:
+            self.conv_depth = nn.Conv2d(self.channels, self.n_bins, kernel_size=1, padding=0, stride=1)
+        else:
+            self.conv_depth = nn.Conv2d(self.channels, 1, kernel_size=1, padding=0, stride=1)
+
+    def _transform_inputs(self, inputs):
+        """Transform inputs for decoder.
+        Args:
+            inputs (list[Tensor]): List of multi-level img features.
+        Returns:
+            Tensor: The transformed inputs
+        """
+
+        if "concat" in self.input_transform:
+            inputs = [inputs[i] for i in self.in_index]
+            if "resize" in self.input_transform:
+                inputs = [
+                    resize(
+                        input=x,
+                        size=[s * self.upsample for s in inputs[0].shape[2:]],
+                        mode="bilinear",
+                        align_corners=self.align_corners,
+                    )
+                    for x in inputs
+                ]
+            inputs = torch.cat(inputs, dim=1)
+        elif self.input_transform == "multiple_select":
+            inputs = [inputs[i] for i in self.in_index]
+        else:
+            inputs = inputs[self.in_index]
+
+        return inputs
+
+    def _forward_feature(self, inputs, img_metas=None, **kwargs):
+        """Forward function for feature maps before classifying each pixel with
+        ``self.cls_seg`` fc.
+        Args:
+            inputs (list[Tensor]): List of multi-level img features.
+        Returns:
+            feats (Tensor): A tensor of shape (batch_size, self.channels,
+                H, W) which is feature map for last layer of decoder head.
+        """
+        # accept lists (for cls token)
+        inputs = list(inputs)
+        for i, x in enumerate(inputs):
+            if len(x) == 2:
+                x, cls_token = x[0], x[1]
+                if len(x.shape) == 2:
+                    x = x[:, :, None, None]
+                cls_token = cls_token[:, :, None, None].expand_as(x)
+                inputs[i] = torch.cat((x, cls_token), 1)
+            else:
+                x = x[0]
+                if len(x.shape) == 2:
+                    x = x[:, :, None, None]
+                inputs[i] = x
+        x = self._transform_inputs(inputs)
+        # feats = self.bn(x)
+        return x
+
+    def forward(self, inputs, img_metas=None, **kwargs):
+        """Forward function."""
+        output = self._forward_feature(inputs, img_metas=img_metas, **kwargs)
+        output = self.depth_pred(output)
+
+        return output