Research map / Detection and segmentation / Tracking
Local All-Pair Correspondence for Point Tracking
Seokju Cho, Jiahui Huang, Jisu Nam, Honggyu An, Seungryong Kim, Joon-Young Lee
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace local 2D correlation with local 4D all-pair correlation (7x7x7x7) (ablation) | LocoTrack-B · TAP-Vid-DAVIS | AJ: 65.0 -> 67.8 | Improved |
| use a dense regular grid of query neighbours instead of random or line-shaped neighbours (ablation) | LocoTrack-B · TAP-Vid-DAVIS | AJ: 65.7 (random) / 66.5 (line) -> 67.2 (grid r_q=1) | Improved |
| enlarge query neighbourhood from 3x3 to 7x7 (ablation) | LocoTrack-B · TAP-Vid-DAVIS | AJ: 67.2 -> 67.8 | Improved |
| use relative position bias instead of sinusoidal position encoding in the Transformer (ablation) | LocoTrack-B · TAP-Vid-DAVIS | AJ: 61.9 -> 67.8 | Improved |
| use a Transformer instead of 1D conv mixer for temporal refinement at equal params (ablation) | LocoTrack-B · TAP-Vid-DAVIS | AJ: 66.1 -> 67.8 | Improved |
| use LocoTrack-S instead of TAPIR on DAVIS strided query | point tracking · TAP-Vid-DAVIS strided (256x256) | AJ: 61.3 -> 66.9 | Improved |
| increase input resolution to 384x512 for LocoTrack | LocoTrack-S · TAP-Vid-RGB-Stacking strided | AJ: 77.4 -> 71.0 | Hurt |
| use LocoTrack-B instead of LocoTrack-S | LocoTrack · TAP-Vid-DAVIS strided (256x256) | AJ: 66.9 -> 67.8 | Improved |
| use LocoTrack-S instead of CoTracker for throughput | point tracking · Throughput | points/sec: 1,146.79 -> 7,244.47 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TAPIR — TAPIR: Tracking Any Point with Per-Frame Initialization and Temporal Refinement (ICCV 2023) · better on 7 of 7 shared results
- CoTracker — CoTracker: It is Better to Track Together (ECCV 2024) · better on 3 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.88)
- Emergent Temporal Correspondences from Video Diffusion Transformers (NeurIPS 2025 · similarity 0.87)
- TAPVid-3D: A Benchmark for Tracking Any Point in 3D (NeurIPS 2024 · similarity 0.86)
- ETAP: Event-based Tracking of Any Point (CVPR 2025 · similarity 0.86)
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · similarity 0.86)
- TrackIME: Enhanced Video Point Tracking via Instance Motion Estimation (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
