why the ResNet backbone not frozen completely?
XiaoBuL opened this issue · 1 comments
XiaoBuL commented
Thanks for your great work.
When I read your code, I find the network optimize the backbone's Layer 2 and Layer 3. Why you do it?
Looking forward to your reply.
DUT-CSJ commented
Thanks for your great work. When I read your code, I find the network optimize the backbone's Layer 2 and Layer 3. Why you do it? Looking forward to your reply.
The SSP model only has a backbone which has learnable parameters. If you freeze it, there is no learnable parameters.