Open rtaori opened 5 years ago
I am seeing a similar issue after re-training the model on the 60 texture dataset. The images output from the training script actually look fine, however the ones from the evaluation script still look like the above. Perhaps it is a bug in the evaluation script or model saving/loading that is causing this problem?
Hi,
I believe there may be an issue with the linked pretrained 60-texture model. I downloaded the linked model and ran the following provided command
th multi_texture_synthesis_test.lua -ind_texture 24
.However, the results look like this: which does not look like the original textures at all.
I have not modified the code at all and followed the intstructions on the readme; does an updated model need to be used here?