Vision Transformer (large-sized model) pre-trained with MSN

Vision Transformer (ViT) model pre-trained using the MSN method. It was introduced in the paper Masked Siamese Networks for Label-Efficient Learning by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas and first released in this repository .

Disclaimer: The team releasing MSN did not write a model card for this model so this model card has been written by the Hugging Face team.

Model description

The Vision Transformer (ViT) is a transformer encoder model (BERT-like). Images are presented to the model as a sequence of fixed-size patches.

MSN presents a joint-embedding architecture to match the prototypes of masked patches with that of the unmasked patches. With this setup, their method yields excellent performance in the low-shot and extreme low-shot regimes.

By pre-training the model, it learns an inner representation of images that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled images for instance, you can train a standard classifier by placing a linear layer on top of the pre-trained encoder.

Intended uses & limitations

You can use the raw model for downstream tasks like image classification. See the model hub to look for different versions of MSN pre-trained models that interest you. The model is particularly beneficial when you have a few labeled samples in your training set.

How to use

Here is how to use this backbone encoder:

from transformers import AutoFeatureExtractor, ViTMSNModel
import torch
from PIL import Image
import requests

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/vit-msn-large")
model = ViTMSNModel.from_pretrained("facebook/vit-msn-large")
inputs = feature_extractor(images=image, return_tensors="pt")
with torch.no_grad():
    outputs = model(**inputs)
last_hidden_states = outputs.last_hidden_state

For fine-tuning on image classification use the ViTMSNForImageClassification class:

from transformers import AutoFeatureExtractor, ViTMSNForImageClassification
import torch
from PIL import Image
import requests

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/vit-msn-large")
model = ViTMSNForImageClassification.from_pretrained("facebook/vit-msn-large")

...

Citation

@article{assran2022masked,
  title={Masked Siamese Networks for Label-Efficient Learning}, 
  author={Assran, Mahmoud, and Caron, Mathilde, and Misra, Ishan, and Bojanowski, Piotr, and Bordes, Florian and Vincent, Pascal, and Joulin, Armand, and Rabbat, Michael, and Ballas, Nicolas},
  journal={arXiv preprint arXiv:2204.07141},
  year={2022}
}