nielsr HF staff commited on
Commit
7213b3a
1 Parent(s): e15d3f3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +32 -27
README.md CHANGED
@@ -1,27 +1,32 @@
1
  ---
2
- license: apache-2.0
3
  tags:
4
  - vision
5
  - image-segmentation
6
  datasets:
7
  - coco
 
 
 
 
 
8
  ---
9
 
10
  # MaskFormer
11
 
12
- MaskFormer model trained on COCO. It was introduced in the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) and first released in [this repository](https://github.com/facebookresearch/MaskFormer/blob/da3e60d85fdeedcb31476b5edd7d328826ce56cc/mask_former/modeling/criterion.py#L169).
13
 
14
- Disclaimer: The team releasing Mask did not write a model card for this model so this model card has been written by the Hugging Face team.
15
 
16
  ## Model description
17
 
18
- MaskFormer addresses semantic segmentation with a mask classification paradigm instead.
19
 
20
  ![model image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/maskformer_architecture.png)
21
 
22
  ## Intended uses & limitations
23
 
24
- You can use the raw model for image classification. See the [model hub](https://huggingface.co/models?search=maskformer) to look for
25
  fine-tuned versions on a task that interests you.
26
 
27
  ### How to use
@@ -29,28 +34,28 @@ fine-tuned versions on a task that interests you.
29
  Here is how to use this model:
30
 
31
  ```python
32
- >>> from transformers import MaskFormerFeatureExtractor, MaskFormerForInstanceSegmentation
33
- >>> from PIL import Image
34
- >>> import requests
35
-
36
- >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
37
- >>> image = Image.open(requests.get(url, stream=True).raw)
38
- >>> feature_extractor = MaskFormerFeatureExtractor.from_pretrained("facebook/maskformer-swin-base-ade")
39
- >>> inputs = feature_extractor(images=image, return_tensors="pt")
40
-
41
- >>> model = MaskFormerForInstanceSegmentation.from_pretrained("facebook/maskformer-swin-base-ade")
42
- >>> outputs = model(**inputs)
43
- >>> # model predicts class_queries_logits of shape `(batch_size, num_queries)`
44
- >>> # and masks_queries_logits of shape `(batch_size, num_queries, height, width)`
45
- >>> class_queries_logits = outputs.class_queries_logits
46
- >>> masks_queries_logits = outputs.masks_queries_logits
47
-
48
- >>> # you can pass them to feature_extractor for postprocessing
49
- >>> output = feature_extractor.post_process_segmentation(outputs)
50
- >>> output = feature_extractor.post_process_semantic_segmentation(outputs)
51
- >>> output = feature_extractor.post_process_panoptic_segmentation(outputs)
 
 
52
  ```
53
 
54
-
55
-
56
  For more code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/master/en/model_doc/maskformer).
 
1
  ---
2
+ license: other
3
  tags:
4
  - vision
5
  - image-segmentation
6
  datasets:
7
  - coco
8
+ widget:
9
+ - src: http://images.cocodataset.org/val2017/000000039769.jpg
10
+ example_title: Cats
11
+ - src: http://images.cocodataset.org/val2017/000000039770.jpg
12
+ example_title: Castle
13
  ---
14
 
15
  # MaskFormer
16
 
17
+ MaskFormer model trained on COCO panoptic segmentation (tiny-sized version, Swin backbone). It was introduced in the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) and first released in [this repository](https://github.com/facebookresearch/MaskFormer/blob/da3e60d85fdeedcb31476b5edd7d328826ce56cc/mask_former/modeling/criterion.py#L169).
18
 
19
+ Disclaimer: The team releasing MaskFormer did not write a model card for this model so this model card has been written by the Hugging Face team.
20
 
21
  ## Model description
22
 
23
+ MaskFormer addresses instance, semantic and panoptic segmentation with the same paradigm: by predicting a set of masks and corresponding labels. Hence, all 3 tasks are treated as if they were instance segmentation.
24
 
25
  ![model image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/maskformer_architecture.png)
26
 
27
  ## Intended uses & limitations
28
 
29
+ You can use this particular checkpoint for semantic segmentation. See the [model hub](https://huggingface.co/models?search=maskformer) to look for other
30
  fine-tuned versions on a task that interests you.
31
 
32
  ### How to use
 
34
  Here is how to use this model:
35
 
36
  ```python
37
+ from transformers import MaskFormerFeatureExtractor, MaskFormerForInstanceSegmentation
38
+ from PIL import Image
39
+ import requests
40
+
41
+ # load MaskFormer fine-tuned on COCO panoptic segmentation
42
+ feature_extractor = MaskFormerFeatureExtractor.from_pretrained("facebook/maskformer-swin-tiny-coco")
43
+ model = MaskFormerForInstanceSegmentation.from_pretrained("facebook/maskformer-swin-tiny-coco")
44
+
45
+ url = "http://images.cocodataset.org/val2017/000000039769.jpg"
46
+ image = Image.open(requests.get(url, stream=True).raw)
47
+ inputs = feature_extractor(images=image, return_tensors="pt")
48
+
49
+ outputs = model(**inputs)
50
+ # model predicts class_queries_logits of shape `(batch_size, num_queries)`
51
+ # and masks_queries_logits of shape `(batch_size, num_queries, height, width)`
52
+ class_queries_logits = outputs.class_queries_logits
53
+ masks_queries_logits = outputs.masks_queries_logits
54
+
55
+ # you can pass them to feature_extractor for postprocessing
56
+ result = feature_extractor.post_process_panoptic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]
57
+ # we refer to the demo notebooks for visualization (see "Resources" section in the MaskFormer docs)
58
+ predicted_panoptic_map = result["segmentation"]
59
  ```
60
 
 
 
61
  For more code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/master/en/model_doc/maskformer).