Ubastic

I'm interested in the hand crafted interaction objective, but I don't see any code related. This objective is more intuitive, can you share it?

I want to implement this in pytorch, is it possible? Cause it is related to 'meta gradient', I tried to implement it but failed. How to achieve this?

could you release training protocols and codes of faster rcnn detection? I'm trying to change this part but the source of detection feature trace back to your work.

how to train the second stage of your agent? Do evaluation and load the best network and train metamemorygraph model? I see there are many related parameter in command parser....

由于实验中需要修改这部分，在论文中提到了codes and pretrained model will be released for reproducibility