Object detection: how to pass the proposals of selective search to Detectron project?

I want to repeat the original fast-rcnn performance with Detectron. Since it does not provide selective search within its code, I obtained proposals of selective search using code here.

According to issue here, I could convert these proposals to some file that Detectron requires using the tool file: proposals->tool file-> some intermediate file->Detectron.

My question is: what parameters should I give to the tool file and Detectron RESPECTIVELY? If possible, examples are preferred.

Thank you in advance.

Sorry for the late reply. I tried to create some code to reproduce the issue.
Would it be possible to provide your proposals, so that we could work with your proposals and some dummy data to debug this issue?

Hi @ptrblck, thank you for your kind reply. To some extent, my problem was solved at this issue.
You can try it using pairs of images and corresponding proposals at here. Thank you sincerely again.

1 Like

I’m glad it’s working now! :slight_smile: