Jun-Yan Zhu
Jun-Yan Zhu
Not sure. We only tested the system on Ubuntu.
@AyushExel any thoughts?
Richard's repo has some pre-trained colorization checkpoints: https://github.com/richzhang/colorization-pytorch
Unfortunately, we didn't store discriminators. Only generators are available.
This might be relevant: https://github.com/broadinstitute/cromwell/issues/3647
It should take 1-2 days on a GTX 1080 GPU. I expect the training time to be much shorter as the GPUs have been much faster compared to 2017.
I am not sure if changing hyper-parameters will help your case. Depending on the type of data you are working on, you might need a new reconstruction loss rather than...
We use Canny edge detector and [HED](https://arxiv.org/abs/1504.06375). We used a similar [script](https://github.com/lllyasviel/ControlNet/tree/main/annotator) as provided by ControlNet.
When you mentioned "it didn't work", does it mean a) the program crashes or 2) the results look worse? 1. For CycleGAN, Resnet-based generators often work much better than UNet....
Sure. It might be worth trying.