Research map / Detection and segmentation / Tracking
Toward Low-Cost yet Effective Temporal Learning for UAV Tracking
Chaocan Xue, Qihua Liang, Bineng Zhong, Yanting Zu, Yuanliang Xue, Haiying Xia, Shuxiang Song
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 |
|---|---|---|---|
| use both attention and classification scores for token extraction instead of attention only (ablation) | LETL token extractor · UAVTrack112_L | AUC: 67.8 -> 68.2 | Improved |
| use both attention and classification scores instead of classification only (ablation) | LETL token extractor · UAVTrack112_L | AUC: 67.4 -> 68.2 | Improved |
| replace merged token value with more recent token instead of averaging (ablation) | LETL token combiner · UAVTrack112_L | AUC: 67.1 -> 68.2 | Improved |
| increase extracted token number K from 16 to 32 (ablation) | LETL · UAVTrack112_L | AUC: 68.2 -> 68.8 (PPF 55 -> 35) | Mixed |
| reduce extracted tokens K from 16 to 8 (ablation) | LETL · UAVTrack112_L | AUC: 67.5 -> 65.2 (P=50%) | Hurt |
| raise merging ratio P from 37.5% to 50% (ablation) | LETL · UAVTrack112_L | AUC: 68.2 -> 67.5 | Hurt |
| lower merging ratio P from 37.5% to 25% (ablation) | LETL · UAVTrack112_L | AUC: 68.2 -> 68.0 | Hurt |
| add LETL temporal module to baseline one-stream tracker (ablation) | DeiT-Tiny one-stream baseline · UAVTrack112_L | precision: +5.0 precision for +0.09 FLOPs | Improved |
| add dense template sequence sampling (ODTrack-style) instead of LETL (ablation) | DeiT-Tiny one-stream baseline · UAVTrack112_L | PPF: +5.9 prec, +0.68 FLOPs, PPF 8.7 | Hurt |
| add dynamic template (MixFormer-style) instead of LETL (ablation) | DeiT-Tiny one-stream baseline · UAVTrack112_L | precision: +3.8 precision, +0.40 FLOPs, PPF 9.5 | Improved |
| add autoregressive queries (AQATrack-style) instead of LETL (ablation) | DeiT-Tiny one-stream baseline · UAVTrack112_L | precision: +4.3 precision, +0.51 FLOPs, PPF 8.4 | Improved |
Closest papers by meaning
- PVT++: A Simple End-to-End Latency-Aware Visual Tracking Framework (ICCV 2023 · similarity 0.89)
- LVPTrack: High Performance Domain Adaptive UAV Tracking with Label Aligned Visual Prompt Tuning (AAAI 2025 · similarity 0.86)
- Dynamic Semantic-Aware Correlation Modeling for UAV Tracking (NeurIPS 2025 · similarity 0.85)
- Learning Occlusion-Robust Vision Transformers for Real-Time UAV Tracking (CVPR 2025 · similarity 0.85)
- Learning Adaptive and View-Invariant Vision Transformer for Real-Time UAV Tracking (ICML 2024 · similarity 0.85)
- Robust Tracking via Mamba-based Context-aware Token Learning (AAAI 2025 · similarity 0.84)
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.
