Has the author tried other U-Net-related decoders, such as U-Net++ or Mamba-UNet?

[Lei0829]

Has the author tried other U-Net-related decoders, such as U-Net++ or Mamba-UNet?

[xiongxyowo]

Hi, we have tried some other decoder designs (such as EMCAD [*1]), but the performance improvement so far is limited. Perhaps using more advanced decoders on some more challenging datasets, coupled with better adapter designs (see issue #37), could better improve performance.
[*1] EMCAD: Efficient Multi-scale Convolutional Attention Decoding for Medical Image Segmentation

---

同学你好，我们尝试过一些其他解码器设计(如EMCAD [*1])，但是目前的性能提升比较有限。也许在一些更具挑战性的数据集上使用更先进的解码器并辅以更好的adapter设计(见 #37 )能获得性能提升。

[Lei0829]

Hi, we have tried some other decoder designs (such as EMCAD [*1]), but the performance improvement so far is limited. Perhaps using more advanced decoders on some more challenging datasets, coupled with better adapter designs (see issue #37), could better improve performance. [*1] EMCAD: Efficient Multi-scale Convolutional Attention Decoding for Medical Image Segmentation

同学你好，我们尝试过一些其他解码器设计(如EMCAD [*1])，但是目前的性能提升比较有限。也许在一些更具挑战性的数据集上使用更先进的解码器并辅以更好的adapter设计(见 #37 )能获得性能提升。

Hello, author. Do you still have the TensorBoard code you used? I’ve been trying to use the code provided by GPT-4, but it keeps throwing errors that I can’t resolve.

[xiongxyowo]

Hi, you may refer to a custom implementation: https://github.com/z520yu/Visual-Cognitive-Engineering-Major-Project/blob/main/train9.py