ACGAN can generate the corresponding target image according to the tag, and its discriminator acts as the tag identification and authenticity judgment in this process. Can a good ACGAN discriminator directly perform multi-classification tasks based on test data sets?If it can't, is there a discriminator in others kind of GAN that can do it? Or need it to add a secondary classifier?
Yes, it is feasible and similar things have already been done.
Here is the link to my Bachelor's thesis where I did the same working on medical images with WGANs and followed by that is the link to a 2020 paper that heavily inspired my thesis:
Thesis Deep neural network with GANs pre-training for tuberculosis ...
https://asset-pdf.scinapse.io/prod/2988141759/2988141759.pdf
- Badges
- Science topic