Views
No views yet
1from urllib.request import urlopen
2from PIL import Image
3import timm
4
5img = Image.open(
6 urlopen('https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'))
7
8model = timm.create_model('efficientformerv2_s2.snap_dist_in1k', pretrained=True)
9model = model.eval()
10
11# get model specific transforms (normalization, resize)
12data_config = timm.data.resolve_model_data_config(model)
13transforms = timm.data.create_transform(**data_config, is_training=False)
14
15output = model(transforms(img).unsqueeze(0)) # unsqueeze single image into batch of 1
16
17top5_probabilities, top5_class_indices = torch.topk(output.softmax(dim=1) * 100, k=5)1from urllib.request import urlopen
2from PIL import Image
3import timm
4
5img = Image.open(
6 urlopen('https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'))
7
8model = timm.create_model(
9 'efficientformerv2_s2.snap_dist_in1k',
10 pretrained=True,
11 num_classes=0, # remove classifier nn.Linear
12)
13model = model.eval()
14
15# get model specific transforms (normalization, resize)
16data_config = timm.data.resolve_model_data_config(model)
17transforms = timm.data.create_transform(**data_config, is_training=False)
18
19output = model(transforms(img).unsqueeze(0)) # output is (batch_size, num_features) shaped tensor
20
21# or equivalently (without needing to set num_classes=0)
22
23output = model.forward_features(transforms(img).unsqueeze(0))
24# output is unpooled (ie.e a (batch_size, num_features, H, W) tensor
25
26output = model.forward_head(output, pre_logits=True)
27# output is (batch_size, num_features) tensor1from urllib.request import urlopen
2from PIL import Image
3import timm
4
5img = Image.open(
6 urlopen('https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'))
7
8model = timm.create_model(
9 'efficientformerv2_s2.snap_dist_in1k',
10 pretrained=True,
11 features_only=True,
12)
13model = model.eval()
14
15# get model specific transforms (normalization, resize)
16data_config = timm.data.resolve_model_data_config(model)
17transforms = timm.data.create_transform(**data_config, is_training=False)
18
19output = model(transforms(img).unsqueeze(0)) # unsqueeze single image into batch of 1
20
21for o in output:
22 # print shape of each feature map in output
23 # e.g. for efficientformerv2_l:
24 # torch.Size([2, 40, 56, 56])
25 # torch.Size([2, 80, 28, 28])
26 # torch.Size([2, 192, 14, 14])
27 # torch.Size([2, 384, 7, 7])
28 print(o.shape)| model | top1 | top5 | param_count | img_size |
|---|---|---|---|---|
| efficientformerv2_l.snap_dist_in1k | 83.628 | 96.54 | 26.32 | 224 |
| efficientformer_l7.snap_dist_in1k | 83.368 | 96.534 | 82.23 | 224 |
| efficientformer_l3.snap_dist_in1k | 82.572 | 96.24 | 31.41 | 224 |
| efficientformerv2_s2.snap_dist_in1k | 82.128 | 95.902 | 12.71 | 224 |
| efficientformer_l1.snap_dist_in1k | 80.496 | 94.984 | 12.29 | 224 |
| efficientformerv2_s1.snap_dist_in1k | 79.698 | 94.698 | 6.19 | 224 |
| efficientformerv2_s0.snap_dist_in1k | 76.026 | 92.77 | 3.6 | 224 |
1@article{li2022rethinking,
2 title={Rethinking Vision Transformers for MobileNet Size and Speed},
3 author={Li, Yanyu and Hu, Ju and Wen, Yang and Evangelidis, Georgios and Salahi, Kamyar and Wang, Yanzhi and Tulyakov, Sergey and Ren, Jian},
4 journal={arXiv preprint arXiv:2212.08059},
5 year={2022}
6}1@misc{rw2019timm,
2 author = {Ross Wightman},
3 title = {PyTorch Image Models},
4 year = {2019},
5 publisher = {GitHub},
6 journal = {GitHub repository},
7 doi = {10.5281/zenodo.4414861},
8 howpublished = {\url{https://github.com/rwightman/pytorch-image-models}}
9}