Segmenter - Transformer for Semantic Segmentation

Segmenter: Transformer for Semantic Segmentation by Robin Strudel, Ricardo Garcia, Ivan Laptev and Cordelia Schmid.

Installation

The code and several trained models will be released soon.

Video Segmentation

Segmentation maps of Seg-B-Mask/16 trained on ADE20K segmentation dataset and tested on DAVIS video dataset.

BibTex

@article{strudel2021,
  title={Segmenter: Transformer for Semantic Segmentation},
  author={Strudel, Robin and Garcia, Ricardo and Laptev, Ivan and Schmid, Cordelia},
  journal={arXiv preprint arXiv:?},
  year={2021}
}

Credits

The Vision Transformer code is based on timm library and the semantic segmentation training and evaluation pipeline is using mmsegmentation.