boahk/media_cdarl

Official repository for "C-DARL: Contrastive diffusion adversarial representation learning for label-free blood vessel segmentation"

MEDIA_CDARL

Official repository for "C-DARL: Contrastive diffusion adversarial representation learning for label-free blood vessel segmentation" published in Medical Image Analysis.

[arXiv][Medical Image Analysis]

Requirements

• OS : Ubuntu
• Python >= 3.9
• PyTorch >= 1.12.1

Data

In our experiments, we used the publicly available XCAD dataset. Please refer to our main paper.

Training

To train our model, run this command:

python3 main.py -p train -c config/train.json

Test

To test the trained our model, run:

python3 main.py -p test -c config/test.json

Pre-trained Models

You can download our pre-trained model of the XCAD dataset here. Then, you can test the model by saving the pre-trained weights in the directory ./experiments/pretrained_model. To briefly test our method given the pre-trained model, we provided the toy example in the directory './data/'.

Citations

@article{kim2024cdarl,
title = {C-DARL: Contrastive diffusion adversarial representation learning for label-free blood vessel segmentation},
journal = {Medical Image Analysis},
volume = {91},
pages = {103022},
year = {2024},
issn = {1361-8415},
doi = {https://doi.org/10.1016/j.media.2023.103022},
author = {Boah Kim and Yujin Oh and Bradford J. Wood and Ronald M. Summers and Jong Chul Ye}
}

media_cdarl's Issues

Unexpected Key Error while resuming training

I am facing an error using your model while resuming training at any epoch, if i do strict=False in the same it says AttributeError: 'CDARL' object has no attribute 'optH'
Could you provide any resolution of it
in load_network
network.load_state_dict(torch.load(H_path), strict=(not self.opt['model']['finetune_norm']))
File "/opt/conda/envs/cdarl/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1667, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for PatchSampleF:
Unexpected key(s) in state_dict: "mlp_0.0.weight", "mlp_0.0.bias", "mlp_0.2.weight", "mlp_0.2.bias", "mlp_1.0.weight", "mlp_1.0.bias", "mlp_1.2.weight", "mlp_1.2.bias", "mlp_2.0.weight", "mlp_2.0.bias", "mlp_2.2.weight", "mlp_2.2.bias".