MOTIP2: Spatial Priors for End-to-End Multi-Object Tracking
End-to-end multi-object trackers have narrowed the gap with classical tracking-by-detection on association-difficult benchmarks.
ProofPaper ↗
Key points
- A model could learn to avoid them, but tracking annotations are scarce, so we encode spatial priors explicitly instead, while keeping inference fully end-to-end with no post-hoc association.
- We propose three spatial priors, at the data, loss, and representation stages.
- Spatial ID Switches bias trajectory permutations toward spatially overlapping objects, reducing the mismatch between training and inference confusions.
- We instantiate the three priors in MOTIP2, a tracker adapted from MOTIP and built on the real-time DEIM detection transformer.
Sources (1)
- [1]MOTIP2: Spatial Priors for End-to-End Multi-Object TrackingarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 04:47 PM
End-to-end multi-object trackers have narrowed the gap with classical tracking-by-detection on association-difficult benchmarks.
A model could learn to avoid them, but tracking annotations are scarce, so we encode spatial priors explicitly instead, while keeping inference fully end-to-end with no post-hoc association.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 7, 2026Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning
- Oct 6, 2026openbmb/MiniCPM-V-4.7-35B-A3B
- Oct 6, 2026EmbeddingGemma 2: an open, lightweight multimodal embedding model
- Oct 5, 2026LiquidAI/d1-omni-600M
- Oct 5, 2026MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers
- Sep 30, 2026Cloudflare/clef-flash