search

run-youngjoo / SC-FEGAN

SC-FEGAN : Face Editing Generative Adversarial Network with User's Sketch and Color (ICCV2019)
Other
3.35k stars 530 forks source link

Add training code #21

Open Abhigupta4981 opened 5 years ago

Abhigupta4981 commented 5 years ago

Training code is necessary to understand what loss functions were used, what was the model architecture etc. It would be a nice idea to add training code too.

abtExp commented 5 years ago

I'm trying to implement it in keras. Training code would be helpful.

Zealoe commented 5 years ago

when will the training codes be released? Thanks.

abtExp commented 5 years ago

i mostly have completed the code with the loss functions and the model, but i'm facing some issues with the model architecture. If anyone can help i'll be able to completely reproduce the model in keras. I've already created a question on keras google group and stack overflow, but still no luck. If anyone of you have any ideas regarding my specific issue, feel free to answer. Here's the link to the question: StackOverflow

claudetheboof commented 5 years ago

Man we really need a training code. Or at least a very well trained model. But training script would be super awesome!

594cp commented 5 years ago

@abtE Thanks for your trying on the training code, do you have any solutions now?

abtExp commented 5 years ago

i'll update the repository this weekend. i still haven't recieved any response on the question, but i'll try to figure it out and will be working on it this weekend. Hopefully i'll get it. 🤞

claudetheboof commented 5 years ago

ugh still no luck in training codes?

Janekaka commented 5 years ago

@abtExp Hi! First of all, thank you for your generosity. May I ask you about your recent progress?

abtExp commented 5 years ago

@claudetheboof @Janekaka and everyone else, sorry for the delay. I've been very busy recently working on my final project submission for my graduation degree and simultaneously working on 2 major projects in my internship. I'll try my best to finish this this month, in the mean time, if anyone of you guys want, you can go ahead, fork my repo or submit a merge request if you can fix some thing or can implement something. Sorry again for the delay and thanks for your patience.

Janekaka commented 5 years ago

Sorry to bother you and thanks for your quick reply. Wish you the best.

abtExp commented 5 years ago

Hi all, finally got some time to work on the implementation, i've fixed the loss functions and the gated convolutions, also the data loader uses the masking technique mentioned in the paper. Although i haven't tested it much, but i think i'll soon complete it and train it. Check out the code at my repo

muxgt commented 4 years ago

@abtExp have you deleted your repository? the link is not available and I haven't found it in the list of your repositories. Any chance to see your code?

abtExp commented 4 years ago

@muxgt hi, sorry for that, I've changed the repository name and structure, I'm developing it now as a library. You can find the repository at : https://github.com/abtExp/arxivr

muxgt commented 4 years ago

@abtExp thanks! I will check it out. Did you receive good results with training?

PngYng commented 4 years ago

@muxgt嗨，很抱歉，我已经更改了存储库名称和结构，现在正在将其开发为一个库。您可以在以下位置找到存储库：https : //github.com/abtExp/arxivr

hi，can you share the train code with me, thanks