onion
onion
Hi, I have reimplemented the blending function, it may help. See https://gist.github.com/onion-liu/7969bce0d4aeb60455f7d35bb0d8fd4a
Hi, I have reimplemented the blending function, it may help. See https://gist.github.com/onion-liu/7969bce0d4aeb60455f7d35bb0d8fd4a
I have tested it on pytorch1.3 + cuda10, it runs successfully
@Chen0403 Hi, Chen, I find that there is an option at SfM->More Functions->Use Shared Calibration, and the result seems better. But I don't know whether it is the right way.
The script [style_transfer_folder.py](https://github.com/onion-liu/BlendGAN/blob/main/style_transfer_folder.py) requires more GPU memory than [generate_image_pairs.py](https://github.com/onion-liu/BlendGAN/blob/main/generate_image_pairs.py) because it loads another model (PSPEncoder). Perhaps you need another GPU with more memory to run style_transfer_folder.py You can also split...
> Yes. But you need to train the new version of the PSPEncoder by yourself.
@dedoogong what is IF method?