Closed koke2c95 closed 3 years ago
Seems interesting, I would need to read up more before implementing something like this. What would be your use case?
Thanks for being interested. First ,I did a lot of searching work and it's none interpretable works on anime dataset at all.
The most common use is style transfer and image transfer more stable image synthesize, like deepdream,feature_vis this may very helpful for these task for handcraft model design! in anime datasets
AND it alse be helpful to do interpretable on the dataset danbooru2019 dataset has 3.69m images with 108m tag without deep cleanup, deep design(imagenet), just from danbooru
RELATED WORKS
DeepDanbooru gradcam
openai microscope pytorch lucid
flashtorch
pytorch-cnn-visualizations Pokemon Grad-CAM