README.md

---
license: mit
tags:
- vision
- image-segmentation
datasets:
- huggan/cityscapes
widget:
- src: https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/cityscapes.png
  example_title: Cityscapes
---

# OneFormer

OneFormer model trained on the Cityscapes dataset (large-sized version, Dinat backbone). It was introduced in the paper [OneFormer: One Transformer to Rule Universal Image Segmentation](https://arxiv.org/abs/2211.06220) by Jain et al. and first released in [this repository](https://github.com/SHI-Labs/OneFormer).

![model image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/oneformer_teaser.png)

## Model description

OneFormer is the first multi-task universal image segmentation framework. It needs to be trained only once with a single universal architecture, a single model, and on a single dataset, to outperform existing specialized models across semantic, instance, and panoptic segmentation tasks. OneFormer uses a task token to condition the model on the task in focus, making the architecture task-guided for training, and task-dynamic for inference, all with a single model.

![model image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/oneformer_architecture.png)

## Intended uses & limitations

You can use this particular checkpoint for semantic, instance and panoptic segmentation. See the [model hub](https://huggingface.co/models?search=oneformer) to look for other fine-tuned versions on a different dataset.

### How to use

Here is how to use this model:

```python
from transformers import OneFormerImageProcessor, OneFormerForUniversalSegmentation
from PIL import Image
import requests
url = "https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/cityscapes.png"
image = Image.open(requests.get(url, stream=True).raw)

# Loading a single model for all three tasks
image_processor = OneFormerImageProcessor.from_pretrained("shi-labs/oneformer_cityscapes_dinat_large")
model = OneFormerForUniversalSegmentation.from_pretrained("shi-labs/oneformer_cityscapes_dinat_large")

# Semantic Segmentation
semantic_inputs = image_processor(images=image, ["semantic"] return_tensors="pt")
semantic_outputs = model(**semantic_inputs)
# pass through image_processor for postprocessing
predicted_semantic_map = image_processor.post_process_semantic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]

# Instance Segmentation
instance_inputs = image_processor(images=image, ["instance"] return_tensors="pt")
instance_outputs = model(**instance_inputs)
# pass through image_processor for postprocessing
predicted_instance_map = image_processor.post_process_instance_segmentation(outputs, target_sizes=[image.size[::-1]])[0]["segmentation"]

# Panoptic Segmentation
panoptic_inputs = image_processor(images=image, ["panoptic"] return_tensors="pt")
panoptic_outputs = model(**panoptic_inputs)
# pass through image_processor for postprocessing
predicted_semantic_map = image_processor.post_process_panoptic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]["segmentation"]
```

For more examples, please refer to the [documentation](https://huggingface.co/docs/transformers/master/en/model_doc/oneformer).

### Citation

```bibtex
@article{jain2022oneformer,
      title={{OneFormer: One Transformer to Rule Universal Image Segmentation}},
      author={Jitesh Jain and Jiachen Li and MangTik Chiu and Ali Hassani and Nikita Orlov and Humphrey Shi},
      journal={arXiv}, 
      year={2022}
    }
```