Research map / Detection and segmentation / Tracking
Distractor-Aware Fast Tracking via Dynamic Convolutions and MOT Philosophy
Zikai Zhang, Bineng Zhong, Shengping Zhang, Zhenjun Tang, Xin Liu, Zhaoxiang Zhang
Check 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 cross-correlation, depthwise cross-correlation, or Hadamard product with dynamic convolutions (ablation) | DMTrack detection correlation · LaSOT | Success: 49.8, 51.4, 51.1 -> 53.0 | Improved |
| integrate dynamic convolutions across layers 1 through 6 rather than only layer 6 (ablation) | DMTrack detection head · LaSOT | Success: 48.7 -> 53.2 | Improved |
| use dense dynamic-convolution integration across layers 1 through 6 instead of the selected 1+6 integration (ablation) | DMTrack detection head · LaSOT | Success: 53.0 -> 53.2 | No clear effect |
| add re-id embedding association rather than position constraint (ablation) | DMTrack object association · LaSOT | Success: 46.9 -> 57.4 | Improved |
| add re-id embedding association rather than appearance constraint (ablation) | DMTrack object association · LaSOT | Success: 55.2 -> 57.4 | Improved |
| use re-id association rather than position constraint (ablation) | DMTrack object association · LaSOT | FPS: 32 -> 31 | Improved |
| use re-id association rather than appearance constraint (ablation) | DMTrack object association · LaSOT | FPS: 16 -> 31 | Improved |
| combine global detection with explicit object association | DMTrack · VOT2018LT | F-score: 0.629 -> 0.683 | Improved |
| use DMTrack instead of Siam R-CNN | long-term tracking · OxUvA test | MaxGM: 72.3 -> 68.8 | Hurt |
| use DMTrack instead of TLD | long-term tracking · OxUvA test | TNR: 89.5 -> 69.4 | Hurt |
| use DMTrack instead of LT_DSE | long-term tracking · VOT2019LT | F-score: 0.695 -> 0.687 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Siam R-CNN — Siam R-CNN: Visual Tracking by Re-Detection (CVPR 2020) · better on 0 of 2 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
- A Distractor-Aware Memory for Visual Object Tracking with SAM2 (CVPR 2025 · similarity 0.90)
- Distractor-Aware Memory-Based Visual Object Tracking (IJCV 2026 · similarity 0.90)
- Distractor-aware Siamese Networks for Visual Object Tracking (ECCV 2018 · similarity 0.89)
- GlobalTrack: A Simple and Strong Baseline for Long-Term Tracking (AAAI 2020 · similarity 0.88)
- Dual-Path Temporal Decoder for End-to-End Multi-Object Tracking (NeurIPS 2025 · similarity 0.88)
- Beyond Local Search: Tracking Objects Everywhere With Instance-Specific Proposals (CVPR 2016 · similarity 0.88)
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.
