phantng / Mask2Former

Implementation of Mask2Former in TensorFlow.

Mask2Former: Masked-attention Mask Transformer for Universal Image Segmentation in TensorFlow 2

Bowen Cheng, Ishan Misra, Alexander G. Schwing, Alexander Kirillov, Rohit Girdhar [arXiv]

Features

• A single architecture for three tasks: panoptic, instance and semantic segmentation. This straightforward mini project was built as part of the main project, IST: A TensorFlow 2 compatible instance segmentation toolbox, with the purpose of adapting recent research into segmentation approaches into TensorFlow.
• Support common benchmark datasets: ADE20K, Cityscapes, COCO, Mapillary Vistas.

Getting started

Project is currently being built, with SwinTransformerV1 and SwinTransformerV2 and a few bits and pieces ready.

License

Shield:

The majority of MaskFormer is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

However portions of the project are available under separate license terms: Swin-Transformer-Semantic-Segmentation is licensed under the MIT license.

Citation

@article{cheng2021mask2former,
  title={Masked-attention Mask Transformer for Universal Image Segmentation},
  author={Bowen Cheng and Ishan Misra and Alexander G. Schwing and Alexander Kirillov and Rohit Girdhar},
  journal={arXiv},
  year={2021}
}