Closed mpapadomanolaki closed 11 months ago
@mpapadomanolaki thanks for your question. In our experiments, we optimize the change probability generator while keeping the VGG feature extractor fixed. But, you can fine-tune the VGG feature extractor as well using a single optimizer.
Thank you for your quick answer!
Hello,
Thank you for your feedback. I am trying to understand your method and I have a question. In your paper you say that the weights of both D-CPG and D-FE are optimized during training. Could you tell me your approach on this? Are you using two different optimizers?
Thank you for your time.
Maria