search

tamarott / SinGAN

Official pytorch implementation of the paper: "SinGAN: Learning a Generative Model from a Single Natural Image"
https://tamarott.github.io/SinGAN.htm
Other
3.31k stars 612 forks source link

Random samples may be generating similar output to input? #50

Open anthony-chaudhary opened 4 years ago

anthony-chaudhary commented 4 years ago

When trying to use random_samples I get results that are appear very similar to the input at most generation scales. 47

Except gen scale of 0 9

Not sure if this is expected or not Also maybe that's just a bad test image / pattern for this type of use? I guess mentally I was thinking it would re arrange the apples or something? It does do that sort of using other function I sent image here: https://github.com/tamarott/SinGAN/issues/2#issuecomment-562814036

I don't know if this is useful or not but I tossed the models and stuff into folder here: https://drive.google.com/open?id=1xe1nbplb0O2hGOt5Y4sFLcjm0UfLtodf

tamarott commented 4 years ago

Hi, The results you are getting seems responsible. However, if you want to get higher changes you can do some "advance experiments" and play with the following parameters: --min_size - you can slightly increase this parameter (let's say to 30) --scale factor - you can slightly decrease this parameter Another option for increasing variability in the image border is to train the model with "noise padding" (see our supplementary material for explanation). This is what happen when training the model for the animation task.