Research map / 3D vision / 3D generation
OccuFly: A 3D Vision Benchmark for Semantic Scene Completion from the Aerial Perspective
Markus Gross, Sai B. Matha, Aya Fahmy, Rui Song, Daniel Cremers, Henri Meeß
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 |
|---|---|---|---|
| fine-tune DepthAnything2 on OccuFly depth maps | DepthAnything2 · OccuFly depth test | delta1: 0.002 -> 0.834 | Improved |
| fine-tune DepthAnything2 on OccuFly depth maps | DepthAnything2 · OccuFly depth test | RMSE: 28.382 -> 4.844 | Improved |
| fine-tune Metric3Dv2 on OccuFly depth maps | Metric3Dv2 · OccuFly depth test | AbsRel: 0.471 -> 0.381 | Improved |
| fine-tune on OccuFly | DepthAnything2 · OccuFly depth test | SILog: 0.192 -> 0.112 (table) vs text: slight degradation | Mixed |
| use SfM+MVS reconstruction instead of DepthAnything3 foundation model | 3D reconstruction · OccuFly scenes | metric scale: 526 % deviation vs 1.24 px reprojection error | Improved |
| use 2D foundation-model pseudo-labels instead of manual annotation | semantic labels · OccuFly | mIoU: 17.58 % mIoU vs manual | Hurt |
| annotate <10% of images and lift to 3D | label transfer · OccuFly | point coverage: >99% points labeled from <10% images | Improved |
| increase altitude from 50 m to 40 m (ablation) | DISC · OccuFly SSC test | IoU: 31.10 -> 27.85 | Hurt |
| use DISC instead of Symphonies | SSC · OccuFly SSC test | IoU: 13.68 -> 29.52 | Improved |
| apply terrestrial SSC models to aerial data | DISC/Symphonies · OccuFly SSC test | mIoU: 2.04 and 0.58 mIoU | Hurt |
Closest papers by meaning
- SDGOCC: Semantic and Depth-Guided Bird's-Eye View Transformation for 3D Multimodal Occupancy Prediction (CVPR 2025 · similarity 0.89)
- Semi-supervised 3D Semantic Scene Completion with 2D Vision Foundation Model Guidance (AAAI 2025 · similarity 0.88)
- IPFormer: Visual 3D Panoptic Scene Completion with Context-Adaptive Instance Proposals (NeurIPS 2025 · similarity 0.88)
- CSV-Occ: Fusing Multi-frame Alignment for Occupancy Prediction with Temporal Cross State Space Model and Central Voting Mechanism (ICML 2025 · similarity 0.88)
- Learning Temporal 3D Semantic Scene Completion via Optical Flow Guidance (NeurIPS 2025 · similarity 0.88)
- H2GFormer: Horizontal-to-Global Voxel Transformer for 3D Semantic Scene Completion (AAAI 2024 · similarity 0.88)
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.
