The Impact of Detector Architecture, Model Scale, and Scene Density on Multi-Pedestrian Tracking

Key points

• Pedestrian tracking in video is important for applications such as surveillance and scene analysis, but it becomes more challenging in crowded scenarios with frequent occlusions and rapid motion.1
• We compare the YOLO11, RT-DETR, and YOLO26 detector families with the ByteTrack and BoT-SORT trackers on sequences from the MOT17, MOT20, and SOMPT22 benchmarks.1
• For YOLO26, we also compare traditional overlap filtering via Non-Maximum Suppression (NMS) with end-to-end inference.1
• The RT-DETR family achieved the highest mean scores on the HOTA tracking quality metric across all three benchmarks with both trackers, with a more pronounced advantage on MOT20, albeit at a higher computational cost.1

Sources (1)

• [1]The Impact of Detector Architecture, Model Scale, and Scene Density on Multi-Pedestrian Tracking
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 01:12 AM

  This paper analyzes how detector architecture, model scale, scene density, and inference regime affect the tracking of multiple pedestrians.

  Pedestrian tracking in video is important for applications such as surveillance and scene analysis, but it becomes more challenging in crowded scenarios with frequent occlusions and rapid motion.

Extractive summary: sentences quoted from the sources.