Research map / Detection and segmentation / CNN features
TAPTR: Tracking Any Point with Transformers as Detection
Hongyang Li, Hao Zhang, Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Lei Zhang
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 |
|---|---|---|---|
| lower the positional-encoding temperature in point self-attention (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 61.9 -> 63.0 | Improved |
| add self attention among point queries (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 58.4 -> 60.9 | Improved |
| add transformer encoder to image features (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 60.9 -> 61.9 | Improved |
| add temporal attention and window feature updating (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 51.6 -> 58.4 | Improved |
| add cost volume aggregation in decoder (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 46.8 -> 51.6 | Improved |
| add cross attention to supplement cost volume (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 50.0 -> 51.6 | Improved |
| use residual content-feature updating instead of DETR-style updating (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 45.1 -> 46.8 | Improved |
| random drop of feature updating in training and gapped feature padding at inference, versus always updating (ablation) | TAPTR · TAP-Vid-RGB-Stacking | AJ: 23.1 -> 60.8 | Improved |
| update cost volume once per window instead of every decoder iteration (ablation) | TAPTR · TAP-Vid-RGB-Stacking | AJ: 58.0 -> 60.8 | Improved |
| update cost volume per window versus never updating (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 63.0 -> 63.0 | No clear effect |
| supervise all decoder layers instead of last only (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 55.4 -> 63.0 | Improved |
| increase decoder layers from 2 to 6 (ablation) | TAPTR · TAP-Vid-DAVIS | AJ: 58.2 -> 63.0 | 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 2 of 3 shared results
- CoTracker — CoTracker: It is Better to Track Together (ECCV 2024) · better on 1 of 1 shared result
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
- TAPTRv2: Attention-based Position Update Improves Tracking Any Point (NeurIPS 2024 · similarity 0.91)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.87)
- TAPVid-3D: A Benchmark for Tracking Any Point in 3D (NeurIPS 2024 · similarity 0.87)
- TAPIP3D: Tracking Any Point in Persistent 3D Geometry (NeurIPS 2025 · similarity 0.87)
- ETAP: Event-based Tracking of Any Point (CVPR 2025 · similarity 0.87)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.85)
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.
