entire codes

[beihaiyou]

Dear Linkai Peng I'm so interested in your work DCDA. But the the codes lack the the key part 'segmentation_models_pytorch',when I download your code in github. I would be so grateful if you give me your entire codes. Looking forward to hearing from you

[lkpengcs]

Thank you for your interest! You can install the library via this link.

[beihaiyou]

Thank you so much for your reply, I will try it right now.

[wh-isper]

Thank you so much for your reply, I will try it right now.

hello，did you reimplement the results?

[lkpengcs]

It may not work on such datasets.

[wh-isper]

I used the polyp dataset, Kvasir, ETIS, and CVC-ColonDB. And I experienced several times, but none of them converge.

[Djoels]

hello, I tried to reproduce results on the OCTA-500 dataset but without success yet. can you provide the samples that were mentioned here?:

"We manually select and form a subset of OCTA-500 (191 from 3M and 254 from 6M), eliminating samples with severe quality issues on either the OCT domain or the OCTA domain. To be noted, the OCT and OCTA images in our subset are unpaired. We resize each image to 384 × 384. In this subset, 50 images are used for testing and the others are used for training."

[lkpengcs]

You can access the dataset via this link.

[Djoels]

thank you very much! will give this a go

[Doris116]

Dear Linkai Peng I am very interested in the work you have done, but the reproduction failed, and there was an error of missing parameters. Another point is how is your dataset directory arranged?

[lkpengcs]

You need to manually set some required parameters. The arrangement of the dataset is the same as the one in the above link.

[Doris116]

After configuring the parameters, an error is reported, indicating that the file is not found. How should I configure the parameters?

[lkpengcs]

Could you check if the path is correct?

[Doris116]

geq各位前辈，请问复现过程中报这种错误是为什么呢，应该如何修改呀

[lkpengcs]

Input channel of the image should be 1

[Doris116]

Thank you for your reply, however, it is strange why this error is reported again after changing the number of input channels to 1. There is a feeling of contradicting each other.

[lkpengcs]

In options.py, you need to set input_dim_a and input_dim_b to 1, not simply reshape the tensor.

[Doris116]

![Uploading c948f69f277422edaea6b9e16204949.png…]() Thanks for the pointers, I encountered another error later on, is this error due to the wrong version? I changed the model.py from "a+=b" to "a= a+b" and it still doesn't work, is there a parameter that needs to be changed?

[lkpengcs]

The picture you attached cannot be seen.