For the part of pretraing, you mentioned in your paper that it is a regular supervised classification task and it learns parameters of backbone which fixed afterwards.
But in the pretrain process, you still use the whole network for training and you fix the parameters of backbone or you only take the backbone for training?
Hi, thanks for providing such an awesome project!
For the part of pretraing, you mentioned in your paper that it is a regular supervised classification task and it learns parameters of backbone which fixed afterwards.
But in the pretrain process, you still use the whole network for training and you fix the parameters of backbone or you only take the backbone for training?
Thanks in advance for your response!