search

yuanming-hu / exposure

Learning infinite-resolution image processing with GAN and RL from unpaired image datasets, using a differentiable photo editing model.
MIT License
767 stars 156 forks source link

How can I train the model using my own data? #13

Open cchen156 opened 6 years ago

cchen156 commented 6 years ago

For example, I have two datasets of PNG images. How can I use the code to train the model, which will transfer the style of one dataset to the other one (unpaired)?

Thank you!

yuanming-hu commented 6 years ago

Hi,

Sorry for the long delay. I get absorbed into several other projects.

Please do a "git pull" and "git submodule update". You can train using images in folders using the new FolderDataProvider. Take a look at the end of config_sintel.py as an example. To train using the Sintel dataset, do "python3 train.py sintel test"

Hope it's still helpful. Btw, I could not come up a good example without copyright issue to demonstrate this usage. If you know any dataset that I can use as a demo (without copyright issues) for such user provided inputs, please let me know... I think professional photos can be good inputs but the copyright issues are making me crazy...

Best, Yuanming

cchen156 commented 6 years ago

Hi Yuanming,

Thank you for your update! For dataset, you are free to use our data: https://github.com/cchen156/Learning-to-See-in-the-Dark

Thank you!

yuanming-hu commented 6 years ago

Thanks! They are interesting data and I will have a try when possible.