megvii-research / video_analyst

A series of basic algorithms that are useful for video understanding, including Single Object Tracking (SOT), Video Object Segmentation (VOS) and so on.
MIT License
832 stars 176 forks source link

About the code in track_head.py #132

Closed SmileTAT closed 4 years ago

SmileTAT commented 4 years ago

Why are the offsets transformed by 'torch.exp(self.si offsets + self.bi) self.total_stride'.Can't find ths in the paper

lzx1413 commented 4 years ago

@SmileTAT This transform is adopted to scale the regress scope for better optimization.