Webssl Mae1b Full2b 224

🚀 Web-SSL MAE ViT-1B: 2B MetaCLIP data, 224 Resolution

This is a 1 billion parameter Vision Transformer (ViT) trained with Masked Autoencoder (MAE) self - supervised learning on web - scale image data without language supervision. It was introduced in "Scaling Language - Free Visual Representation Learning" (Fan et al., 2025).

🚀 Quick Start

The Web - SSL MAE 1B model is a powerful tool for vision tasks. You can start using it by following the steps in the "Usage Examples" section.

✨ Features

• Trained on 2 billion web images without language supervision, demonstrating the power of pure visual learning.
• Can match or exceed the performance of language - supervised models like CLIP across various vision tasks.
• Shows particularly strong performance on OCR & Chart understanding tasks and maintains competitive performance in traditional vision benchmarks and multimodal tasks.

📦 Installation

The installation process mainly involves installing the transformers library. You can use the following command:

pip install transformers

💻 Usage Examples

Basic Usage

from transformers import AutoImageProcessor, ViTModel
import torch
from PIL import Image

# Adjust the size, crop_size, etc. fields to your liking
processor = AutoImageProcessor.from_pretrained('facebook/webssl-mae1b-full2b-224')
model = ViTModel.from_pretrained('facebook/webssl-mae1b-full2b-224').cuda().eval()

# Process an image
image = Image.open('path/to/image.jpg')
inputs = processor(images=image, return_tensors="pt").to('cuda')
with torch.no_grad():
    outputs = model(**inputs)

# Extract features from the encoder
encoder_hidden_states = outputs.last_hidden_state

📚 Documentation

Model Details

Model Descriptions

Web - SSL MAE 1B is a 1 billion parameter Vision Transformer model. It is trained using masked autoencoder self - supervised learning on 2 billion web images without language supervision. This model shows that when pure visual learning is scaled appropriately, it can achieve comparable or better performance than language - supervised models like CLIP in various vision tasks. Web - MAE performs especially well in OCR & Chart understanding tasks and remains competitive in traditional vision benchmarks and multimodal tasks.

📄 License

This project is licensed under the cc - by - nc - 4.0 license.

📚 Citation

@article{fan2025scaling,
  title={Scaling Language-Free Visual Representation Learning}, 
  author={David Fan and Shengbang Tong and Jiachen Zhu and Koustuv Sinha and Zhuang Liu and Xinlei Chen and Michael Rabbat and Nicolas Ballas and Yann LeCun and Amir Bar and Saining Xie},
  year={2025},
  eprint={2504.01017},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}