search

mit-han-lab / gan-compression

[CVPR 2020] GAN Compression: Efficient Architectures for Interactive Conditional GANs
Other
1.1k stars 148 forks source link

Gray-Scale Input Support #100

Closed ailen-wrx closed 2 years ago

ailen-wrx commented 2 years ago

Hello, thanks for your great work! I am currently training your network on my own dataset with gray-scale images but the code, to be specific, metric/fid_score.py and metric/inception.py failed to execute correctly. Gray-scale input is not considered when computing fid_score in metric/fid_score.py, and the pretrained inception_v3 model does not support single-channel input.

I made a few modification on the code on my forked repo to make it run on my dataset: commit

Still I do not know whether my code works well or not. Hoping that you will let me know if official support is provided :)

lmxyy commented 2 years ago

Could you just convert the gray scale images to the RGB ones? FID calculation does not support single channel. Also, you could duplicate the channels three times, but I am not sure if the FID number is reliable.