Research map / 3D vision / Human pose estimation
PiLoT: Neural Pixel-to-3D Registration for UAV-based Ego and Target Geo-localization
Xiaoya Cheng, Long Wang, Yan Liu, Xinyi Liu, Hanlin Tan, Yu Liu, Maojun Zhang, Shen Yan
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 domain-specific training on synthetic UAV data to off-the-shelf backbone (ablation) | PiLoT · Ablation, 10m/10deg init | Recall@1m/1deg (%): 0.0 -> 15.2 | Improved |
| add rotation-aware hypothesis generation (ablation) | PiLoT · Ablation, 10m/10deg init | Recall@1m/1deg (%): 15.2 -> 70.6 | Improved |
| add motion regularizer for hypothesis selection (ablation) | PiLoT · Ablation, 10m/10deg init | Recall@1m/1deg (%): 70.6 -> 84.2 | Improved |
| train on synthetic data with light/weather instead of without (ablation) | PiLoT · Ablation, 3m/3deg init | Recall@1m/1deg (%): 63.5 -> 84.3 | Improved |
| train on synthetic data instead of MegaDepth only (ablation) | PiLoT · Ablation, 3m/3deg init | Recall@1m/1deg (%): 69.9 -> 84.3 | Improved |
| use PiLoT instead of Render2Loc (LoFTR) | localization baseline · UAVD4L-2yr (zero-shot) | median translation error (m): 1.08 -> 0.92 | Improved |
| use PiLoT instead of best baseline on SynthCity-6 | Render2Loc (RoMaV2) · SynthCity-6 | median translation error (m): 0.47 -> 0.46 | Improved |
| use PiLoT instead of Render2Loc for target geo-localization | Render2Loc · single-target real | Recall@1m (%): 87.62 -> 90.81 | Improved |
| use PiLoT for target geo-localization in multi-target synthetic | Render2Loc · multi-target synthetic | Recall@1m (%): 89.03 -> 93.74 | Improved |
| use PiLoT instead of Render2Loc (LoFTR) at R@1 on UAVScenes | Render2Loc (LoFTR) · UAVScenes (zero-shot) | Recall@1m/1deg (%): 23.2 -> 25.5 | Mixed |
Closest papers by meaning
- Game4Loc: A UAV Geo-Localization Benchmark from Game Data (AAAI 2025 · similarity 0.90)
- Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery (NeurIPS 2026 · similarity 0.89)
- LoD-Loc: Aerial Visual Localization using LoD 3D Map with Neural Wireframe Alignment (NeurIPS 2024 · similarity 0.89)
- Seeing Across Skies and Streets: Feedforward 3D Reconstruction from Satellite, Drone, and Ground Images (NeurIPS 2026 · similarity 0.87)
- SOAR: Regression-based LiDAR Relocalization for UAVs (NeurIPS 2026 · similarity 0.87)
- OrthoLoC: UAV 6-DoF Localization and Calibration Using Orthographic Geodata (NeurIPS 2025 · similarity 0.87)
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.
