Vision Transformer (ViT)

The Vision Transformer is a model for image classification that employs a Transformer-like architecture over patches of the image. This includes the use of Multi-Head Attention, Scaled Dot-Product Attention and other architectural features seen in the Transformer architecture traditionally used for NLP.

How do I use this model on an image?

To load a pretrained model:

import timm
model = timm.create_model('vit_base_patch16_224', pretrained=True)
model.eval()

To load and preprocess the image:

import urllib
from PIL import Image
from timm.data import resolve_data_config
from timm.data.transforms_factory import create_transform

config = resolve_data_config({}, model=model)
transform = create_transform(**config)

url, filename = ("https://github.com/pytorch/hub/raw/master/images/dog.jpg", "dog.jpg")
urllib.request.urlretrieve(url, filename)
img = Image.open(filename).convert('RGB')
tensor = transform(img).unsqueeze(0) # transform and add batch dimension

To get the model predictions:

import torch
with torch.no_grad():
    out = model(tensor)
probabilities = torch.nn.functional.softmax(out[0], dim=0)
print(probabilities.shape)
# prints: torch.Size([1000])

To get the top-5 predictions class names:

# Get imagenet class mappings
url, filename = ("https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt", "imagenet_classes.txt")
urllib.request.urlretrieve(url, filename) 
with open("imagenet_classes.txt", "r") as f:
    categories = [s.strip() for s in f.readlines()]

# Print top categories per image
top5_prob, top5_catid = torch.topk(probabilities, 5)
for i in range(top5_prob.size(0)):
    print(categories[top5_catid[i]], top5_prob[i].item())
# prints class names and probabilities like:
# [('Samoyed', 0.6425196528434753), ('Pomeranian', 0.04062102362513542), ('keeshond', 0.03186424449086189), ('white wolf', 0.01739676296710968), ('Eskimo dog', 0.011717947199940681)]

Replace the model name with the variant you want to use, e.g. vit_base_patch16_224. You can find the IDs in the model summaries at the top of this page.

To extract image features with this model, follow the timm feature extraction examples, just change the name of the model you want to use.

How do I finetune this model?

You can finetune any of the pre-trained models just by changing the classifier (the last layer).

model = timm.create_model('vit_base_patch16_224', pretrained=True, num_classes=NUM_FINETUNE_CLASSES)

To finetune on your own dataset, you have to write a training loop or adapt timm's training script to use your dataset.

How do I train this model?

You can follow the timm recipe scripts for training a new model afresh.

Citation

@misc{dosovitskiy2020image,
      title={An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale}, 
      author={Alexey Dosovitskiy and Lucas Beyer and Alexander Kolesnikov and Dirk Weissenborn and Xiaohua Zhai and Thomas Unterthiner and Mostafa Dehghani and Matthias Minderer and Georg Heigold and Sylvain Gelly and Jakob Uszkoreit and Neil Houlsby},
      year={2020},
      eprint={2010.11929},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}