Research map / 3D vision / Depth and stereo
SemCityLoc: Aerial 6DoF Localization Using Semantic 3D City Models
Jingfeng Mao, Xuyang Chen, Qilin Zhang, Oussema Dhaouadi, Guangming Wang, Brian Sheil, Daniel Cremers, Yan Xia, Olaf Wysocki
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 |
|---|---|---|---|
| replace LoD-Loc with SemCityLoc semantic-depth alignment | LoD-Loc · SemCityLockeD | recall @2m-2deg (%): 35.11 -> 69.15 | Improved |
| add coarse semantic cost-volume selection (ablation) | SemCityLoc · SemCityLockeD | recall @2m-2deg (%): 41.49 -> 69.15 | Improved |
| add semantic-depth particle-filter refinement (ablation) | SemCityLoc · SemCityLockeD | recall @2m-2deg (%): 56.38 -> 69.15 | Improved |
| add coarse selection (ablation) | SemCityLoc · UAVD4L-LoD in-Traj | recall @2m-2deg (%): 62.03 -> 70.45 | Improved |
| add refinement (ablation) | SemCityLoc · UAVD4L-LoD in-Traj | recall @2m-2deg (%): 47.07 -> 70.45 | Improved |
| SemCityLoc vs LoD-Loc | SemCityLoc · UAVD4L-LoD in-Traj | recall @2m-2deg (%): 83.42 -> 70.45 | Hurt |
| SemCityLoc vs LoD-Loc | LoD-Loc · Swiss-EPFL out-of-Place | mean XYZ error (m): 12.54 -> 3.09 | Improved |
| SemCityLoc vs LoD-Loc | LoD-Loc · Swiss-EPFL in-Place | recall @2m-2deg (%): 36.23 -> 33.99 | Mixed |
| use LoD3 instead of LoD2 model (ablation) | SemCityLoc · SemCityLockeD | recall @3m-3deg (%): 84.04 -> 75.53 | No clear effect |
| use LoD1 instead of LoD2 model (ablation) | SemCityLoc · SemCityLockeD | recall @2m-2deg (%): 69.15 -> 48.94 | Hurt |
| use zero-shot CLIP+Semantic-SAM segmentation instead of trained DINOv3 head (ablation) | SemCityLoc · SemCityLockeD | recall @2m-2deg (%): 69.15 -> 15.93 | Hurt |
| increase prior translation perturbation from 10-20 m to 150-200 m (ablation) | SemCityLoc · SemCityLockeD | recall @2m-2deg (%): 68.09 -> 18.09 | Hurt |
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.
- LoD-Loc — LoD-Loc: Aerial Visual Localization using LoD 3D Map with Neural Wireframe Alignment (NeurIPS 2024) · better on 7 of 9 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
- LoD-Loc: Aerial Visual Localization using LoD 3D Map with Neural Wireframe Alignment (NeurIPS 2024 · similarity 0.86)
- OrthoLoC: UAV 6-DoF Localization and Calibration Using Orthographic Geodata (NeurIPS 2025 · similarity 0.86)
- Seeing Across Skies and Streets: Feedforward 3D Reconstruction from Satellite, Drone, and Ground Images (NeurIPS 2026 · similarity 0.85)
- Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery (NeurIPS 2026 · similarity 0.84)
- 360Loc: A Dataset and Benchmark for Omnidirectional Visual Localization with Cross-device Queries (CVPR 2024 · similarity 0.84)
- Aerial Lifting: Neural Urban Semantic and Building Instance Lifting from Aerial Imagery (CVPR 2024 · 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.
