Looking at the Inceptionv3 model defined in PyTorch ( see this file ), is there a way to use its pre-trained version but removing its fully connected layer ? I am not looking to do fine-tuning. I am looking to completely eliminate the fully connected layer as I am trying to do something else.
- I hope this may be helpful
removed = list(Inceptionv3.children())[:-1] Inceptionv3 = nn.Sequential(*self.removed)
and here --> Use ResNet18 as feature extractor is a question like yours
Thanks a lot. It works