Thanks for your awesome reproduce work about partial convolution work, while I made two quick implementation of this repo on two different machines, both of my implementation seem to have some problems about data, specifically, I suspect that the problem caused by flow_from_directory function.
The first error sceenshot looks like:
it says the fit_generator receives empty batch.
its really awkward since I didn't change the code, I just provide the training data path. and the training data organizes like this:
train/
CelebHQ/
001.jpg
002.jpg
...
Any suggestions or discussiones are welcomed!
Thanks for your awesome reproduce work about partial convolution work, while I made two quick implementation of this repo on two different machines, both of my implementation seem to have some problems about data, specifically, I suspect that the problem caused by flow_from_directory function.
The first error sceenshot looks like: it says the fit_generator receives empty batch.
its really awkward since I didn't change the code, I just provide the training data path. and the training data organizes like this: train/ CelebHQ/ 001.jpg 002.jpg ... Any suggestions or discussiones are welcomed!