CaptainEven / MCMOT

Real time one-stage multi-class & multi-object tracking based on anchor-free detection and ReID
MIT License
383 stars 82 forks source link

Possible bug in Flip augmentation #54

Closed austinmw closed 3 years ago

austinmw commented 3 years ago

When the image is flipped along the x-axis, the labels are modified by: https://github.com/CaptainEven/MCMOT/blob/24f2efb943ecafe297a68deb7d10b45c2750894e/src/lib/datasets/dataset/jde.py#L287

Since at this point the labels are in xywh format, this maintains the width which is correct, but I believe the new x1 value should be (1-x-w) instead of (1-x), or:

labels[:, 2] = 1 - (labels[:, 2] + labels[:, 4])

Or am I missing something?