Research map / Detection and segmentation / Tracking
Rethinking Temporal Modeling in Visual Object Tracking via Decoupled Auxiliary Supervision
Dailing Zhang, Shiyu Hu, Honghao Fu, Xiaokun Feng, Yipei Wang, Kang Hao Cheong, Kaiqi Huang
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 |
|---|---|---|---|
| Add decoupled auxiliary supervision (template-masked auxiliary pass) in training (ablation) | HiViT tracker · LaSOT | AUC: 73.0 -> 73.4 | Improved |
| Add decoupled auxiliary supervision (ablation) | ViT tracker · LaSOT | AUC: 71.0 -> 71.4 | Improved |
| Add temporal tokens with standard joint training (ablation) | HiViT tracker · LaSOT | AUC: 72.8 -> 73.0 | No clear effect |
| Perturb temporal tokens at inference | Standard temporal trackers (ARTrackV2, EVPTrack, AQATrack, ODTrack) · LaSOT | AUC: 71.4 -> 71.2 (ARTrackV2) | No clear effect |
| Provide ground-truth cropped template or GT-centered search at inference | SUTrack / OSTrack · LaSOT | AUC: 73.1 -> 82.1; 69.9 -> 77.9 | Improved |
| Give auxiliary branch template access (ablation) | DAS · LaSOT | AUC: 73.4 -> 72.6 | Hurt |
| Use adaptive (learnable-uncertainty) loss weight instead of fixed (ablation) | DAS · LaSOT | AUC: 73.2 -> 73.4 | Improved |
| Use LoRA rank 64 vs no LoRA (ablation) | DAS · LaSOT | AUC: 73.2 -> 73.4 | Improved |
| Increase LoRA rank to 128 (ablation) | DAS · LaSOT | AUC: 73.4 -> 72.7 | Hurt |
| Use joint instead of sequential optimization of auxiliary task (ablation) | DAS · LaSOT | AUC: 72.7 -> 73.4 | Improved |
| Add DAS to training (ablation) | Temporal-token gradient ratio · LaSOT | R_H: 0.165 -> 0.624 | Improved |
| Use DAS (training-only) at inference | DASTrack vs SUTrack · Speed/params | fps: 34.0 -> 34.0 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SUTrack384 — SUTrack: Towards Simple and Unified Single Object Tracking (AAAI 2025) · better on 4 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
- ODTrack: Online Dense Temporal Token Learning for Visual Tracking (AAAI 2024 · similarity 0.88)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.86)
- VastTrack: Vast Category Visual Object Tracking (NeurIPS 2024 · similarity 0.86)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.86)
- A Distractor-Aware Memory for Visual Object Tracking with SAM2 (CVPR 2025 · similarity 0.86)
- Tracktention: Leveraging Point Tracking to Attend Videos Faster and Better (CVPR 2025 · 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.
