reafly/ldnet

MICCAI 2022 : Lesion-aware Dynamic Kernel for Polyp Segmentation (Pytorch implementation).

ldnet's Issues

About pytorch version

I can't run your released code, could you tell me the version of pytorch for your experiment?

ValueError

ValueError: Unknown resampling filter (None). Use Image.Resampling.NEAREST (0), Image.Resampling.BILINEAR (2) or Image.Resampling.BICUBIC (3)

code

Your work is very interesting. When will you release the code. thank you.

I think there may be a fatal mistake in your research

In the code, the indicators of your segmentation results are obtained by converting the GT and prediction results to a resolution of 224 * 224, and the polyp segmentation result graph saved in your test. py is also a resolution of 224 * 224.
It is obvious that the indicators of segmentation results should be obtained by upsampling the predicted results with the same resolution as GT and GT, and the polyp segmentation result image saved in test.py should also have the original resolution of the image.
I cannot accept this calculation method, and others cannot accept it, which is completely unreasonable.

在代码中，你们的分割结果的指标是通过转换到224224的分辨率的ground_truth和预测结果得到的，你们的test.py中保存的息肉分割结果图也是224224的分辨率。
很明显的，分割结果的指标应该由上采样为和ground_truth一样分辨率的预测结果和ground_truth得到，test.py中保存的息肉分割结果图也应该图片原来的分辨率。
我无法接受这样的计算方式，其他人也无法接受，这完全不合理。

Dataset Request

Hello author, I noticed some differences in the way you partition the dataset compared to articles such as PraNet, or in the folder branches of the dataset. Can you share your dataset?