README.md
3.7 KB · 79 lines · markdown Raw
1 ---
2 license: mit
3 tags:
4 - vision
5 - image-segmentation
6 datasets:
7 - scene_parse_150
8 widget:
9 - src: https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/ade20k.jpeg
10 example_title: House
11 - src: https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/demo_2.jpg
12 example_title: Airplane
13 - src: https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/coco.jpeg
14 example_title: Person
15 ---
16
17 # OneFormer
18
19 OneFormer model trained on the ADE20k dataset (tiny-sized version, Swin 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).
20
21 ![model image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/oneformer_teaser.png)
22
23 ## Model description
24
25 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.
26
27 ![model image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/oneformer_architecture.png)
28
29 ## Intended uses & limitations
30
31 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.
32
33 ### How to use
34
35 Here is how to use this model:
36
37 ```python
38 from transformers import OneFormerProcessor, OneFormerForUniversalSegmentation
39 from PIL import Image
40 import requests
41 url = "https://huggingface.co/datasets/shi-labs/oneformer_demo/blob/main/ade20k.jpeg"
42 image = Image.open(requests.get(url, stream=True).raw)
43
44 # Loading a single model for all three tasks
45 processor = OneFormerProcessor.from_pretrained("shi-labs/oneformer_ade20k_swin_tiny")
46 model = OneFormerForUniversalSegmentation.from_pretrained("shi-labs/oneformer_ade20k_swin_tiny")
47
48 # Semantic Segmentation
49 semantic_inputs = processor(images=image, task_inputs=["semantic"], return_tensors="pt")
50 semantic_outputs = model(**semantic_inputs)
51 # pass through image_processor for postprocessing
52 predicted_semantic_map = processor.post_process_semantic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]
53
54 # Instance Segmentation
55 instance_inputs = processor(images=image, task_inputs=["instance"], return_tensors="pt")
56 instance_outputs = model(**instance_inputs)
57 # pass through image_processor for postprocessing
58 predicted_instance_map = processor.post_process_instance_segmentation(outputs, target_sizes=[image.size[::-1]])[0]["segmentation"]
59
60 # Panoptic Segmentation
61 panoptic_inputs = processor(images=image, task_inputs=["panoptic"], return_tensors="pt")
62 panoptic_outputs = model(**panoptic_inputs)
63 # pass through image_processor for postprocessing
64 predicted_semantic_map = processor.post_process_panoptic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]["segmentation"]
65 ```
66
67 For more examples, please refer to the [documentation](https://huggingface.co/docs/transformers/master/en/model_doc/oneformer).
68
69 ### Citation
70
71 ```bibtex
72 @article{jain2022oneformer,
73 title={{OneFormer: One Transformer to Rule Universal Image Segmentation}},
74 author={Jitesh Jain and Jiachen Li and MangTik Chiu and Ali Hassani and Nikita Orlov and Humphrey Shi},
75 journal={arXiv},
76 year={2022}
77 }
78 ```
79