DTMFormer

This repo is the official implementation for:
[AAAI2024] DTMFormer: Dynamic Token Merging for Boosting Transformer-Based Medical Image Segmentation.
(The details of our DTMFormer can be found at the models directory in this repo or in the paper. We take SETR for example.)

Requirements

• python 3.6
• pytorch 1.8.0
• torchvision 0.9.0

Datasets

• The ACDC dataset could be acquired from here.
• The ISIC 2018 dataset could be acquired from here.
• The BTCV dataset could be acquired from here.

Training

Commands for training

python train.py

Testing

Commands for testing

python test.py

References

1. vit-pytorch