nwuai/can

Context-assisted full Attention Network for brain tissue segmentation

can's Issues

Model checkpoints problem

I'm trying to use the pretrained weights. But the pretrained model return zero dice score. I checked the weights and found that it seems to be the first epoch saved model instead of the final model.
Could you please provide the pretrained model you used in your article?

Looking forward to your reply.

dHCP dataset split.

Hi, thank you for your contributiuon.

Can you share the datasplit of dHCP2017 dataset in origin sample id format? The preprocessed dataset has ordered id. The origin dataset has sample id like CC00069XX12.

Looking forward to your reply.

In testing stage, the input of the used network is different from the input of training stage.

Hello, for the released code by you, I found the input of the network is different from the input of the testing stage when reproducing your segmentation result, i.e., there are image_3D,skull_3D,img_2D in the training stage, while there are only image_3D in the testing stage. The released code is right?

Error while testing for dhcp dataset

I am trying to run test.py for dhcp dataset from the pretrained weights. I am getting following error,

Namespace(workers=6, resume='X:\DPC-4\CAN-main\segmentation\save_model\dhcp\checkpoint_pretrain.pth.tar', save_dir='\save_model\dhcp_2', data_dir='..\Datasets\dhcp', data_list='data_list', encode3D=True, se_loss=False, use_weights=False, epochs=100, start_epoch=0, test_batch_size=1, num_slices=5, num_class=10, cuda=True, seed=1, lr=0.01, lr_scheduler='poly', weight_decay=0.0001)
Traceback (most recent call last):
File "X:\DPC-4\CAN-main\segmentation\test.py", line 259, in
main()
File "X:\DPC-4\CAN-main\segmentation\test.py", line 254, in main
solver = Solver(args, args.num_class)
File "X:\DPC-4\CAN-main\segmentation\test.py", line 84, in init
self.optimizer.load_state_dict(checkpoint['optimizer'])
File "C:\anaconda\lib\site-packages\torch\optim\optimizer.py", line 385, in load_state_dict
raise ValueError("loaded state dict has a different number of "
ValueError: loaded state dict has a different number of parameter groups

What parameters I should change in model for dhcp? I have set num-classes = 10

Source Dataset Unavaliable.

Hi, I want to obtain the source MALC dataset. I apply for the dataset on the official website using the contact form but no reponse is returned.

I would appreciate it if you could share the source dataset or telling me how to apply for the dataset in a right way.