Recently, transformers have been proposed as a promising tool for global context modeling by employing a powerful global attention mechanism, but one of their main shortcomings when applied to segmentation tasks is that they cannot effectively extract sufficient local details to tackle ambiguous boundaries. We propose a novel boundary-aware transformer (BAT) to comprehensively address the challenges of automatic skin lesion segmentation.
This paper has been accepted by MICCAI. Get the full paper on Arxiv.
- Network
- Pre-processing
- Training Codes
- MS
For more details or any questions, please feel easy to contact us by email ^_^
-
First, you can download the dataset from ISIC challenge.
-
Second, for pre-processing the dataset, you can run:
$ python src/resize.py
$ python src/point_gen.py
-
Third, before running the network, you should first download the code of CELL_DETR into lib and install it.
-
In the end, for testing the model, you could run:
$ python net/trans_deeplab.py
We will update the latest training version under the same setting as CA-Net.
If you find BAT useful in your research, please consider citing:
@inproceedings{wang2021boundary,
title={Boundary-Aware Transformers for Skin Lesion Segmentation},
author={Wang, Jiacheng and Wei, Lan and Wang, Liansheng and Zhou, Qichao and Zhu, Lei and Qin, Jing},
booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
pages={206--216},
year={2021},
organization={Springer}
}