Research map / 3D vision / Gaussian splatting
Free-Viewpoint Video of Outdoor Sports Using a Drone
Zhengdong Hong
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 drone motion constraint (DMC) in bundle adjustment (ablation) | drone calibration · AerialRecon 300 trajectories (calibration) | total failures: 37 -> 0 | Improved |
| add sequential feature matching (ablation) | drone calibration · AerialRecon 300 trajectories (calibration) | total failures: 6 -> 0 | Improved |
| use 2D transformer keypoint detector instead of CNN (ablation) | Drone-based sequential MoCap · AerialRecon 25 scenes (pose estimation) | AP: 64.5 -> 67.9 | Improved |
| add temporal smoothness term to SMPL optimization (ablation) | Drone-based sequential MoCap · AerialRecon 25 scenes (pose estimation) | AP: 66.3 -> 67.9 | Improved |
| use orbiting drone footage instead of handheld camera (ablation) | free-viewpoint rendering system · AerialRecon Football | PSNR: 15.34 -> 21.03 | Improved |
| proposed system versus HOSNeRF on its own dataset | Ours · HOSNeRF dataset | PSNR: Tennis 23.95 vs 24.50; Suitcase 21.67 vs 21.79; Lounge 28.91 vs 27.84 | Mixed |
| proposed system versus baselines on AerialRecon sports scenes | Ours · AerialRecon Climb | PSNR: 21.80 -> 24.24 (vs HOSNeRF) | Improved |
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.
- HOSNeRF — HOSNeRF: Dynamic Human-Object-Scene Neural Radiance Fields from a Single Video (ICCV 2023) · better on 6 of 9 shared results
- OpenPose — OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields (TPAMI 2019) · 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
- UAV4D: Dynamic Neural Rendering of Human-Centric UAV Imagery Using Gaussian Splatting (AAAI 2026 · similarity 0.85)
- Learning to Film From Professional Human Motion Videos (CVPR 2019 · similarity 0.85)
- DroneSplat: 3D Gaussian Splatting for Robust 3D Reconstruction from In-the-Wild Drone Imagery (CVPR 2025 · similarity 0.83)
- AerialMegaDepth: Learning Aerial-Ground Reconstruction and View Synthesis (CVPR 2025 · similarity 0.83)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.81)
- UAV3D: A Large-scale 3D Perception Benchmark for Unmanned Aerial Vehicles (NeurIPS 2024 · similarity 0.81)
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.
