fanq15
Two way Contrastive Training
Thanks for your code but I got confused about two way training and wish your help. I can understand 1 way training, the surpport image and quary image feed into network and output boxes with the same class of surpport image. But when it comes to 2 way training, is there 2 corresponding classes with 2 surpport images? if it's just one positive class, how the negtive branch output proposal in rpn? Thanks a lot!
Positive and negative classes have their own attention rpn. The targets of negative class rpn are all 0. But there are still some proposals generated by the negative class rpn and will be sent to the detector to classify them.
Positive and negative classes have their own attention rpn. The targets of negative class rpn are all 0. But there are still some proposals generated by the negative class rpn and will be sent to the detector to classify them.
Does the RPN/Multi-Relation Detector share parameters between 2 branch?