mostafa-saad / segment-correction-net Goto Github PK

Hello author, is your code open source? I've been waiting for a long time, thank you!

Hello author, is your code open source?

I used to do medical image detection, using b-box. Your method can achieve semantic segmentation in a semi supervised way. But I don't have any code. Can you roughly predict whether your method is feasible in medical images based on b-box? thank you.

Hi,
Wonderful idea and work!
I meet some trouble when trying to reproduce the results.
My ancilliary model is built and trained by following the setting in the paper and achieves mIoU=85.77% in VOC validation set. However, when I use the Linear Correction to train the primary model, the performance only achieves around 77%.
I follow the settings in the paper as follows:
\alpha descrease from 30.0 to 0.5.
Learning rate start from 0.007 with poly decay schedule.
Stop the gradient at the refined soft labels.
Any advice?
Thanks~

Hello authors,

Thank you for your paper. I just wanted to ask about code release -- will you release code?

Thank you!

Hello author, is your code open source?