zihangJiang/TokenLabeling
PyTorch implementation of LV-ViT (Large-scale Vision Transformer) with token labeling for improved image classification and segmentation.
Not currently ranked — collecting fresh signals.
star history
This repository implements the paper ‘All Tokens Matter: Token Labeling for Training Better Vision Transformers’. It provides training code and pre-trained models for LV-ViT, a vision transformer architecture trained using a token labeling approach that assigns labels to every token rather than just the CLS token. The implementation is based on the timm (pytorch-image-models) library and includes scripts for label data generation and a segmentation model variant.
Frequently asked
- What is zihangJiang/TokenLabeling?
- PyTorch implementation of LV-ViT (Large-scale Vision Transformer) with token labeling for improved image classification and segmentation.
- Is TokenLabeling open source?
- Yes — zihangJiang/TokenLabeling is open source, released under the Apache-2.0 license.
- What language is TokenLabeling written in?
- zihangJiang/TokenLabeling is primarily written in Jupyter Notebook.
- How popular is TokenLabeling?
- zihangJiang/TokenLabeling has 435 stars on GitHub.
- Where can I find TokenLabeling?
- zihangJiang/TokenLabeling is on GitHub at https://github.com/zihangJiang/TokenLabeling.