search

SharifAmit / RVGAN

[MICCAI'21] [Tensorflow] Retinal Vessel Segmentation using a Novel Multi-scale Generative Adversarial Network
BSD 3-Clause "New" or "Revised" License
88 stars 20 forks source link

How should we train the model using tf 2.6.0? #8

Closed xiaogangzhu closed 3 years ago

xiaogangzhu commented 3 years ago

Hi, I tried training the model on tf 2.6.0 and some problems happens. How should i fix this? 2021-10-20 16-10-57 的屏幕截图

SharifAmit commented 3 years ago

Currently we are not planning to create a branch that supports TF 2.6.0.

However, you can check one of our recent works which utilizes TF 2.6.0, where we used similar Coarse and Fine generators, and Discriminator with a embedding feature-matching loss. The overall training is quite similar, and should be reproducible for RVGAN.

Link

Hope this helps !

xiaogangzhu commented 3 years ago

Is your recent work VTGAN works for vessel segmentation? I want to try the model on my own retinal pictures and it seems there is only pretrained model so I need to train by myself. But unfortunately the scripts don't work for my RTX3090 lol. I was wondering if I can get the final model weight file and it will be really helpful!

SharifAmit commented 3 years ago

I have created a new branch for tf 2.6.0

Link: https://github.com/SharifAmit/RVGAN/tree/tf-2.6

Please open a new issue if you have any error.

Thanks!

xiaogangzhu commented 3 years ago

Thank you!