Robust-and-efficient-post-processing-for-video-object-detection
Robust-and-efficient-post-processing-for-video-object-detection copied to clipboard
ms in recoordinate_tubelets_full of REPP.py
I note that the variance for gaussian smoothing for coordinates is set to 0.6 in the paper. But it is set to self.recoordinate_std*100/ms in this line in this respository. What are the effect of the augment ms and the ratio 100/ms here? Thanks!