nielsr HF staff commited on
Commit
78fe42e
1 Parent(s): 491661a

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +78 -0
README.md ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ tags:
4
+ - vision
5
+ - image-segmentation
6
+ datasets:
7
+ - scene_parse_150
8
+ widget:
9
+ - src: https://huggingface.co/datasets/hf-internal-testing/fixtures_ade20k/resolve/main/ADE_val_00000001.jpg
10
+ example_title: House
11
+ - src: https://huggingface.co/datasets/hf-internal-testing/fixtures_ade20k/resolve/main/ADE_val_00000002.jpg
12
+ example_title: Castle
13
+ ---
14
+
15
+ # SegFormer (b0-sized) encoder fine-tuned on ADE20k
16
+
17
+ SegFormer encoder fine-tuned on Imagenet-1k. It was introduced in the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Xie et al. and first released in [this repository](https://github.com/NVlabs/SegFormer).
18
+
19
+ Disclaimer: The team releasing SegFormer 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
+ SegFormer consists of a hierarchical Transformer encoder and a lightweight all-MLP decode head to achieve great results on semantic segmentation benchmarks such as ADE20K and Cityscapes. The hierarchical Transformer is first pre-trained on ImageNet-1k, after which a decode head is added and fine-tuned altogether on a downstream dataset.
24
+
25
+ This repository only contains the pre-trained hierarchical Transformer, hence it can be used for fine-tuning purposes.
26
+
27
+ ## Intended uses & limitations
28
+
29
+ You can use the model for fine-tuning of semantic segmentation. See the [model hub](https://huggingface.co/models?other=segformer) to look for fine-tuned versions on a task that interests you.
30
+
31
+ ### How to use
32
+
33
+ Here is how to use this model to classify an image of the COCO 2017 dataset into one of the 1,000 ImageNet classes:
34
+
35
+ ```python
36
+ from transformers import SegformerFeatureExtractor, SegformerForImageClassification
37
+ from PIL import Image
38
+ import requests
39
+
40
+ url = "http://images.cocodataset.org/val2017/000000039769.jpg"
41
+ image = Image.open(requests.get(url, stream=True).raw)
42
+
43
+ feature_extractor = SegformerFeatureExtractor.from_pretrained("nvidia/mit-b0")
44
+ model = SegformerForImageClassification.from_pretrained("nvidia/mit-b0")
45
+
46
+ inputs = feature_extractor(images=image, return_tensors="pt")
47
+ outputs = model(**inputs)
48
+ logits = outputs.logits
49
+ # model predicts one of the 1000 ImageNet classes
50
+ predicted_class_idx = logits.argmax(-1).item()
51
+ print("Predicted class:", model.config.id2label[predicted_class_idx])
52
+ ```
53
+
54
+ For more code examples, we refer to the [documentation](https://huggingface.co/transformers/model_doc/segformer.html#).
55
+
56
+ ### BibTeX entry and citation info
57
+
58
+ ```bibtex
59
+ @article{DBLP:journals/corr/abs-2105-15203,
60
+ author = {Enze Xie and
61
+ Wenhai Wang and
62
+ Zhiding Yu and
63
+ Anima Anandkumar and
64
+ Jose M. Alvarez and
65
+ Ping Luo},
66
+ title = {SegFormer: Simple and Efficient Design for Semantic Segmentation with
67
+ Transformers},
68
+ journal = {CoRR},
69
+ volume = {abs/2105.15203},
70
+ year = {2021},
71
+ url = {https://arxiv.org/abs/2105.15203},
72
+ eprinttype = {arXiv},
73
+ eprint = {2105.15203},
74
+ timestamp = {Wed, 02 Jun 2021 11:46:42 +0200},
75
+ biburl = {https://dblp.org/rec/journals/corr/abs-2105-15203.bib},
76
+ bibsource = {dblp computer science bibliography, https://dblp.org}
77
+ }
78
+ ```