question about the recall metric in paper

[jihwanp]

Hi I have a small question about the recall metric you've used in ur paper.

1. In table 1, you report the rpn recall value and in other tables for novel categories, you also report the recall value.
   Maybe you evaluated the former in class-agnostic manner, but did you evaluate the latter class-specific? e.g., considering the class prediction at recall value
2. If the first question is true, have you tried to evaluate the latter in a class-agnostic manner? Im curious if your method (IKD,GKD) can improve the box proposal quality at novel categories. (e.g., if baseline cant propose the box related to the novel categ but yours can, it would be very interesting )

Sorry for the detailed request for you since I'm struggling to use your code to reproduce the results :)