search

infinite-tao / MA-MTLN

Multi-attention Guided Multi-task Learning Network for Automatic Gastric Tumor Segmentation and Lymph Node Classification
34 stars 10 forks source link

Share the existing methods, such as CMSVNet #1

Open chengjianhong opened 3 years ago

chengjianhong commented 3 years ago

Hi, Thank you for your excellent work. Recently, I am studying multi-task learning and compare it with some existing methods. A method, CmsVNet (Zhou et al. 2021), mentioned in your paper is used for comparison. I attempt to implement it, but it works poorly. Maybe my implementation is wrong. Can you share your CmsVNet implementation?

Best wishes, jianhong

infinite-tao commented 3 years ago

Thank you for your affirmation of our work. We have shared the implementation of comparative method CmsVNet.

Best regards, Yongtao

chengjianhong commented 3 years ago

@infinite-tao Hi, thank you for your sharing the implementation. But I find that the backbone network of your implementation is different from the architecture of Zhou et al. 2021. Their proposed network seems very simple, only including several convolution blocks consisting of conv+BN+ReLU. Are you referring to the idea of iteration in his article?

infinite-tao commented 3 years ago

For fair comparison, we employed SENet-50 as the backbone of the network. In addition, we replaced Conv+BN+ReLU with Conv+GN+PReLU.

chengjianhong commented 3 years ago

@infinite-tao OK, Thanks.