deeplearning-wisc / stud

source code for CVPR'22 paper "Unknown-Aware Object Detection: Learning What You Don’t Know from Videos in the Wild"
Apache License 2.0
117 stars 14 forks source link

Confusing about t_0. #9

Closed jinweiisgreat closed 2 years ago

jinweiisgreat commented 2 years ago

Thank you for your nice paper! I have a question about the key frame :

d12306 commented 2 years ago

hi, @jinweiisgreat , thanks for the interest in our work!

For frame t0, it is randomly selected from the frames in one video. we deal with t0 in the same way as in the other frames, i.e., letting it going through the object detection networks. The only different thing from the reference frames is that we force the energy scores for the objects that are in frame t0 to be higher.