2D图像的多类别分割

[fm52Hz]

作者您好，我看到你是基于VM-Unet这个项目，并且在前面的多类别问题中的也是用的VM-Unet。我想问一下，你是否能成功使用VM-Unet训练多类别分割呢？因为我使用那个项目训练发现loss并没有下降，最终的效果也很差。再请问 #54 中把train_synapse.py中的模型换成UltraLight_VM_UNet，再进行一系列的修改吗？感谢！

[wurenkai]

Hi, thank you for your interest in our work. As you said, you can train our UltraLight_VM_UNet for multi-classification tasks with replacement based on VMUNet's project. As you mentioned, VMUNet does have a slow loss decline for both binary and multiclassification tasks, and it usually shows a significant decline only at a certain epoch. This may be influenced by the model architecture of VMUNet itself, specifically, you can consult this author at the VMUNet project.

[Bougaivillea10]

@fm52Hz 您好，请问您更换后是否在synapse上实现了分割效果