Frame-by-Frame Detection Plus BoT-SORT is Stronger

AkshatS07 · x · 2026-07-11

The post mentions that combining the frame-by-frame detect API with BoT-SORT yields better-than-expected performance.

However, the current shortcoming is re-identification: maintaining the identity of the same target across different frames or camera angles isn't quite there yet. The author adds that their video model will attempt to tackle this issue very soon.

Original post →

More from Research

Research channel →