Researchers have developed ARC-SORT, a new multi-object tracking system designed for sports broadcasts that operates in real-time without requiring any training. The system measures its own uncertainty to adapt its tracking behavior, using an Interacting Multiple Model (IMM) Kalman filter to define search regions based on prediction reliability. ARC-SORT achieved high accuracy on the SportsMOT and DanceTrack datasets, outperforming existing methods in terms of tracking accuracy and identity switches, and runs efficiently on consumer hardware. AI
IMPACT This new tracking algorithm could improve automated analysis and fan engagement in sports broadcasts.
RANK_REASON This is a research paper detailing a new algorithm and its performance on benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
- ARC-SORT
- DanceTrack
- Deep-EIoU
- GeForce RTX 3060
- Girish Kumar Adari
- Interacting Multiple Model (IMM) Kalman filter
- SportsMOT
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →