Update README.md
Browse files
README.md
CHANGED
@@ -77,29 +77,32 @@ model-index:
|
|
77 |
pipeline_tag: image-segmentation
|
78 |
---
|
79 |
|
80 |
-
# FLAIR model collection
|
81 |
-
|
82 |
-
The FLAIR models are a collection of semantic segmentation models initially developed to classify land cover on very high resolution aerial images (more specifically the French [BD ORTHO®](https://geoservices.ign.fr/bdortho) product).
|
83 |
-
The distributed pre-trained models differ in their :
|
84 |
-
* dataset for training : [**FLAIR** dataset] (https://huggingface.co/datasets/IGNF/FLAIR) or the increased version of this dataset **FLAIR-INC** (x 3.5 patches). Only the FLAIR dataset is open at the moment.
|
85 |
-
* input modalities : **RGB** (natural colours), **RGBI** (natural colours + infrared), **RGBIE** (natural colours + infrared + elevation)
|
86 |
-
* model architecture : **resnet34_unet** (U-Net with a Resnet-34 encoder), **deeplab**
|
87 |
-
* target class nomenclature : **12cl** (12 land cover classes) or **15cl** (15 land cover classes)
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
# FLAIR FLAIR-INC_RVBIE_resnet34_unet_15cl_norm model
|
92 |
-
|
93 |
-
The general characteristics of this specific model **FLAIR-INC_RVBIE_resnet34_unet_15cl_norm** are :
|
94 |
-
* trained with the FLAIR-INC dataset
|
95 |
-
* RGBIE images (true colours + infrared + elevation)
|
96 |
-
* U-Net with a Resnet-34 encoder
|
97 |
-
* 15 class nomenclature : [building, pervious_ surface, impervious_surface, bare_soil, water, coniferous, deciduous, brushwood, vineyard, herbaceous, agricultural_land, plowed_land, swimming pool, snow, greenhouse]
|
98 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
## Model Informations
|
101 |
-
|
102 |
-
- **Code repository:** https://github.com/IGNF/FLAIR-1-AI-Challenge
|
103 |
- **Paper:** https://arxiv.org/pdf/2211.12979.pdf
|
104 |
- **Developed by:** IGN
|
105 |
- **Compute infrastructure:**
|
@@ -107,7 +110,8 @@ The general characteristics of this specific model **FLAIR-INC_RVBIE_resnet34_un
|
|
107 |
- hardware: HPC/AI resources provided by GENCI-IDRIS
|
108 |
- **License:** : Apache 2.0
|
109 |
|
110 |
-
|
|
|
111 |
## Uses
|
112 |
|
113 |
Although the model can be applied to other type of very high spatial earth observation images, it was initially developed to tackle the problem of classifying aerial images acquired on the French Territory.
|
@@ -148,15 +152,15 @@ _**Using the model on other spatial areas**_ :
|
|
148 |
The FLAIR-INC_RVBIE_resnet34_unet_15cl_norm model was trained on patches reprensenting the French Metropolitan territory.
|
149 |
The user should be aware that applying the model to other type of landscapes may imply a drop in model metrics.
|
150 |
|
151 |
-
|
152 |
|
153 |
## How to Get Started with the Model
|
154 |
|
155 |
-
|
|
|
156 |
|
157 |
-
Use the code below to get started with the model.
|
158 |
-
{{ get_started_code | default("[More Information Needed]", true)}}
|
159 |
|
|
|
160 |
|
161 |
## Training Details
|
162 |
|
@@ -166,14 +170,16 @@ Use the code below to get started with the model.
|
|
166 |
The train/validation split was performed patchwise to obtain a 80% / 20% distribution between train and validation.
|
167 |
Annotation was performed at the _zone_ level (~100 patches per _zone_). Spatial independancy between patches is guaranted as patches from the same _zone_ were assigned to the same set (TRAIN or VALIDATION).
|
168 |
The following number of patches were used for train and validation :
|
169 |
-
| TRAIN set
|
170 |
| VALIDATION set | 43 700 patchs |
|
171 |
|
172 |
|
173 |
-
### Training Procedure
|
174 |
|
175 |
|
176 |
-
|
|
|
|
|
|
|
177 |
|
178 |
For traning the model, input normalization was performed to center-reduce (**a mean=0** and a **standard deviation = 1**, channel wise) the dataset.
|
179 |
We used the statistics of TRAIN+VALIDATION for input normalization. It is recommended that the user apply the same type of input normalization.
|
@@ -189,14 +195,9 @@ Statistics of the TRAIN+VALIDATION set :
|
|
189 |
| Elevation Channel (E) | 53.26 |79.30 |
|
190 |
|
191 |
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
#### Training Hyperparameters
|
196 |
|
197 |
-
|
198 |
-
|
199 |
-
* Model architecture: Unet (implementation from the [Segmentation Models Pytorch library](https://segmentation-modelspytorch.readthedocs.io/en/latest/docs/api.html#unet)
|
200 |
* Encoder : Resnet-34 pre-trained with ImageNet
|
201 |
* Augmentation :
|
202 |
* VerticalFlip(p=0.5)
|
@@ -215,8 +216,7 @@ Statistics of the TRAIN+VALIDATION set :
|
|
215 |
* Class Weights : [1-building: 1.0 , 2-pervious surface: 1.0 , 3-impervious surface: 1.0 , 4-bare soil: 1.0 , 5-water: 1.0 , 6-coniferous: 1.0 , 7-deciduous: 1.0 , 8-brushwood: 1.0 , 9-vineyard: 1.0 , 10-herbaceous vegetation: 1.0 , 11-agricultural land: 1.0 , 12-plowed land: 1.0 , 13-swimming_pool: 1.0 , 14-snow: 1.0 , 15-clear cut: 0.0 , 16-mixed: 0.0 , 17-ligneous: 0.0 , 18-greenhouse: 1.0 , 19-other: 0.0]
|
216 |
|
217 |
|
218 |
-
#### Speeds, Sizes, Times
|
219 |
-
|
220 |
|
221 |
The FLAIR-INC_RVBIE_resnet34_unet_15cl_norm model was trained on a HPC/AI resources provided by GENCI-IDRIS (Grant 2022-A0131013803).
|
222 |
16 V100 GPUs were used ( 4 nodes, 4 GPUS per node). With this configuration the approximate learning time is 6 minutes per epoch.
|
@@ -295,32 +295,35 @@ The following illustration gives the resulting confusion matrix :
|
|
295 |
</div>
|
296 |
|
297 |
|
298 |
-
|
299 |
-
|
300 |
-
|
301 |
### Results
|
302 |
|
303 |
Samples of results
|
304 |
|
305 |
|
306 |
-
|
|
|
|
|
307 |
|
308 |
|
309 |
**BibTeX:**
|
310 |
|
311 |
-
|
|
|
312 |
title={FLAIR: a Country-Scale Land Cover Semantic Segmentation Dataset From Multi-Source Optical Imagery},
|
313 |
author={Anatol Garioud and Nicolas Gonthier and Loic Landrieu and Apolline De Wit and Marion Valette and Marc Poupée and Sébastien Giordano and Boris Wattrelos},
|
314 |
year={2023},
|
315 |
-
|
316 |
-
|
317 |
-
primaryClass={cs.CV}
|
318 |
}
|
|
|
319 |
|
320 |
|
321 |
**APA:**
|
322 |
-
|
323 |
-
|
324 |
-
|
325 |
-
|
326 |
-
|
|
|
|
|
|
|
|
77 |
pipeline_tag: image-segmentation
|
78 |
---
|
79 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
80 |
|
81 |
+
<div style="border:0px; padding:25px; background-color:#F8F5F5; padding-top:10px; padding-bottom:1px;">
|
82 |
+
<h1>FLAIR model collection</h1>
|
83 |
+
<p>The FLAIR models are a collection of semantic segmentation models initially developed to classify land cover on very high resolution aerial images (more specifically the French <a href="https://geoservices.ign.fr/bdortho">BD ORTHO®</a> product). The distributed pre-trained models differ in their :</p>
|
84 |
+
<ul style="list-style-type:disc;">
|
85 |
+
<li>dataset for training : <a href="https://huggingface.co/datasets/IGNF/FLAIR"><b>FLAIR</b> dataset</a> or the increased version of this dataset <b>FLAIR-INC</b> (x 3.5 patches). Only the FLAIR dataset is open at the moment.</li>
|
86 |
+
<li>input modalities : <b>RGB</b> (natural colours), <b>RGBI</b> (natural colours + infrared), <b>RGBIE</b> (natural colours + infrared + elevation)</li>
|
87 |
+
<li>model architecture : <b>resnet34_unet</b> (U-Net with a Resnet-34 encoder), <b>deeplab</b></li>
|
88 |
+
<li>target class nomenclature : <b>12cl</b> (12 land cover classes) or <b>15cl</b> (15 land cover classes)</li>
|
89 |
+
</ul>
|
90 |
+
</div>
|
91 |
+
<br>
|
92 |
+
|
93 |
+
<div style="border:0px; padding:25px; background-color:#FDFFF4 ; padding-top:10px; padding-bottom:1px;">
|
94 |
+
<h1>FLAIR FLAIR-INC_RVBIE_resnet34_unet_15cl_norm model</h1>
|
95 |
+
<p>The general characteristics of this specific model <strong>FLAIR-INC_RVBIE_resnet34_unet_15cl_norm</strong> are :</p>
|
96 |
+
<ul style="list-style-type:disc;">
|
97 |
+
<li>Trained with the FLAIR-INC dataset</li>
|
98 |
+
<li>RGBIE images (true colours + infrared + elevation)</li>
|
99 |
+
<li>U-Net with a Resnet-34 encoder</li>
|
100 |
+
<li>15 class nomenclature : [building, pervious surface, impervious surface, bare soil, water, coniferous, deciduous, brushwood, vineyard, herbaceous, agricultural land, plowed land, swimming pool, snow, greenhouse]</li>
|
101 |
+
</ul>
|
102 |
+
</div>
|
103 |
|
104 |
## Model Informations
|
105 |
+
- **Code repository:** https://github.com/IGNF/FLAIR-1
|
|
|
106 |
- **Paper:** https://arxiv.org/pdf/2211.12979.pdf
|
107 |
- **Developed by:** IGN
|
108 |
- **Compute infrastructure:**
|
|
|
110 |
- hardware: HPC/AI resources provided by GENCI-IDRIS
|
111 |
- **License:** : Apache 2.0
|
112 |
|
113 |
+
---
|
114 |
+
|
115 |
## Uses
|
116 |
|
117 |
Although the model can be applied to other type of very high spatial earth observation images, it was initially developed to tackle the problem of classifying aerial images acquired on the French Territory.
|
|
|
152 |
The FLAIR-INC_RVBIE_resnet34_unet_15cl_norm model was trained on patches reprensenting the French Metropolitan territory.
|
153 |
The user should be aware that applying the model to other type of landscapes may imply a drop in model metrics.
|
154 |
|
155 |
+
---
|
156 |
|
157 |
## How to Get Started with the Model
|
158 |
|
159 |
+
Visit ([https://github.com/IGNF/FLAIR-1](https://github.com/IGNF/FLAIR-1)) to use the model.
|
160 |
+
Fine-tuning and prediction tasks are detailed in the README file.
|
161 |
|
|
|
|
|
162 |
|
163 |
+
---
|
164 |
|
165 |
## Training Details
|
166 |
|
|
|
170 |
The train/validation split was performed patchwise to obtain a 80% / 20% distribution between train and validation.
|
171 |
Annotation was performed at the _zone_ level (~100 patches per _zone_). Spatial independancy between patches is guaranted as patches from the same _zone_ were assigned to the same set (TRAIN or VALIDATION).
|
172 |
The following number of patches were used for train and validation :
|
173 |
+
| TRAIN set | 174 700 patches |
|
174 |
| VALIDATION set | 43 700 patchs |
|
175 |
|
176 |
|
|
|
177 |
|
178 |
|
179 |
+
|
180 |
+
### Training Procedure
|
181 |
+
|
182 |
+
#### Preprocessing
|
183 |
|
184 |
For traning the model, input normalization was performed to center-reduce (**a mean=0** and a **standard deviation = 1**, channel wise) the dataset.
|
185 |
We used the statistics of TRAIN+VALIDATION for input normalization. It is recommended that the user apply the same type of input normalization.
|
|
|
195 |
| Elevation Channel (E) | 53.26 |79.30 |
|
196 |
|
197 |
|
|
|
|
|
|
|
198 |
#### Training Hyperparameters
|
199 |
|
200 |
+
* Model architecture: Unet (implementation from the [Segmentation Models Pytorch library](https://segmentation-modelspytorch.readthedocs.io/en/latest/docs/api.html#unet))
|
|
|
|
|
201 |
* Encoder : Resnet-34 pre-trained with ImageNet
|
202 |
* Augmentation :
|
203 |
* VerticalFlip(p=0.5)
|
|
|
216 |
* Class Weights : [1-building: 1.0 , 2-pervious surface: 1.0 , 3-impervious surface: 1.0 , 4-bare soil: 1.0 , 5-water: 1.0 , 6-coniferous: 1.0 , 7-deciduous: 1.0 , 8-brushwood: 1.0 , 9-vineyard: 1.0 , 10-herbaceous vegetation: 1.0 , 11-agricultural land: 1.0 , 12-plowed land: 1.0 , 13-swimming_pool: 1.0 , 14-snow: 1.0 , 15-clear cut: 0.0 , 16-mixed: 0.0 , 17-ligneous: 0.0 , 18-greenhouse: 1.0 , 19-other: 0.0]
|
217 |
|
218 |
|
219 |
+
#### Speeds, Sizes, Times
|
|
|
220 |
|
221 |
The FLAIR-INC_RVBIE_resnet34_unet_15cl_norm model was trained on a HPC/AI resources provided by GENCI-IDRIS (Grant 2022-A0131013803).
|
222 |
16 V100 GPUs were used ( 4 nodes, 4 GPUS per node). With this configuration the approximate learning time is 6 minutes per epoch.
|
|
|
295 |
</div>
|
296 |
|
297 |
|
|
|
|
|
|
|
298 |
### Results
|
299 |
|
300 |
Samples of results
|
301 |
|
302 |
|
303 |
+
---
|
304 |
+
|
305 |
+
## Citation
|
306 |
|
307 |
|
308 |
**BibTeX:**
|
309 |
|
310 |
+
```
|
311 |
+
@inproceedings{ign-flair,
|
312 |
title={FLAIR: a Country-Scale Land Cover Semantic Segmentation Dataset From Multi-Source Optical Imagery},
|
313 |
author={Anatol Garioud and Nicolas Gonthier and Loic Landrieu and Apolline De Wit and Marion Valette and Marc Poupée and Sébastien Giordano and Boris Wattrelos},
|
314 |
year={2023},
|
315 |
+
booktitle={Advances in Neural Information Processing Systems (NeurIPS) 2023},
|
316 |
+
doi={https://doi.org/10.48550/arXiv.2310.13336},
|
|
|
317 |
}
|
318 |
+
```
|
319 |
|
320 |
|
321 |
**APA:**
|
322 |
+
```
|
323 |
+
Anatol Garioud, Nicolas Gonthier, Loic Landrieu, Apolline De Wit, Marion Valette, Marc Poupée, Sébastien Giordano and Boris Wattrelos. 2023.
|
324 |
+
FLAIR: a Country-Scale Land Cover Semantic Segmentation Dataset From Multi-Source Optical Imagery. (2023).
|
325 |
+
In proceedings of Advances in Neural Information Processing Systems (NeurIPS) 2023.
|
326 |
+
DOI: https://doi.org/10.48550/arXiv.2310.13336
|
327 |
+
```
|
328 |
+
|
329 |
+
## Contact : ai-challenge@ign.fr
|