neural-motifs icon indicating copy to clipboard operation
neural-motifs copied to clipboard

Published 20 hours ago • verified publisher icon rowanz

Questions about RoIAlign

Open qingzwang opened this issue 6 years ago • 2 comments

When I read your code, I am confused by the calculation of RoI align. In RoIAlignFunction, the normalized rois are computed as follows: rois_normalized[:, 1] /= width and width=(data_width - 1) / saptial_scale which should be the width of the input image, so the normalized rois are in [0, 1], is this correct? From the paper of msak-rcnn, I think rois_normalized[:, 1] *= spatial_scale.

qingzwang avatar Feb 21 '19 04:02 qingzwang

Not sure what you mean here. I believe I got the RoIAlign implementation from other implementations of the same idea, though the paper is a bit vague on details so maybe something changed. In both cases you're multiplying by spatial_scale right?

rowanz avatar Mar 11 '19 20:03 rowanz

Dear author, Thank you very much for providing such excellent code, but I always show that bbox_overlaps can't be found in the process of running, but I have generated the. so file. Do you have any suggestions to help me?

L6-hong avatar Oct 12 '20 08:10 L6-hong