Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond the Horizon: Decoupling Multi-View UAV Action Recognition via Partial Order Transfer
Wenxuan Liu, Zhuo Zhou, Xuemei Jia, Siyuan Yang, Wenxin Huang, Xian Zhong, Chia-Wen Lin
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 Action Partial Order Guide (APOG) to X3D (ablation) | X3D baseline · DRONE-ACTION | Top-1 accuracy (%): 83.4 -> 93.1 | Improved |
| add View Discrete Guide (VDG) to X3D (ablation) | X3D baseline · DRONE-ACTION | Top-1 accuracy (%): 83.4 -> 91.7 | Improved |
| combine APOG and VDG (ablation) | X3D baseline · UAV | Top-1 accuracy (%): 32.3 -> 43.2 | Improved |
| transfer knowledge from low-altitude view to high-altitude view (v1 -> v2), 2 views (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 90.2 -> 94.2 | Improved |
| reverse transfer direction to high -> low altitude (v2 -> v1), 2 views (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 90.2 -> 87.5 | Hurt |
| sequential transfer v1 -> v2 -> v3, 3 views (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 91.7 -> 97.4 | Improved |
| reversed sequential transfer v3 -> v2 -> v1, 3 views (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 91.7 -> 88.9 | Hurt |
| transfer from ground NTU-RGB+D view to UAV (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 90.2 -> 95.8 | Improved |
| transfer from UAV to ground NTU-RGB+D (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 90.2 -> 88.9 | Hurt |
| use three view partitions instead of two, four or five (ablation) | Aerorder · DRONE-ACTION, MOD20, UAV | Top-1 accuracy (%) | Improved |
| set guidance loss weight to 10 instead of 0.1 (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 97.4 -> 91.7 | Hurt |
| set decoupling loss weight to 0.01 instead of 1 (ablation) | Aerorder · DRONE-ACTION | Top-1 accuracy (%): 97.4 -> 93.1 | Hurt |
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.
- ASAT — ASAT: Adaptive Scoring and Thresholding with Human Feedback for Robust Out-of-Distribution Detection (NeurIPS 2026) · better on 1 of 1 shared result
- DVANet — DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 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
- Multiview Aerial Visual RECognition (MAVREC): Can Multi-view Improve Aerial Visual Perception? (CVPR 2024 · similarity 0.87)
- 3D-aware Select, Expand, and Squeeze Token for Aerial Action Recognition (AAAI 2025 · similarity 0.87)
- Towards Realistic UAV Vision-Language Navigation: Platform, Benchmark, and Methodology (ICLR 2025 · similarity 0.85)
- WorldVLN: Autoregressive World Action Model for Aerial Vision-Language Navigation (NeurIPS 2026 · similarity 0.85)
- Drones Help Drones: A Collaborative Framework for Multi-Drone Object Trajectory Prediction and Beyond (NeurIPS 2024 · similarity 0.84)
- Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery (NeurIPS 2026 · 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.
