Research map / 3D vision / 3D generation
SkyScenes: A Synthetic Dataset for Aerial Scene Understanding
Sahil S Khose, Anisha Pal, Aayushi Agarwal, . Deepanshi, Judy Hoffman, Prithvijit Chattopadhyay
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 |
|---|---|---|---|
| train on SkyScenes instead of SynDrone | DAFormer (MiT-B5) · AeroScapes | mIoU: 30.53 -> 40.72 | Improved |
| train on SkyScenes instead of SynDrone | Rein (DINOv2) · UAVid | mIoU: 54.92 -> 54.19 | No clear effect |
| train on SkyScenes instead of SynDrone | Rein (DINOv2) · ICG Drone person IoU | IoU: 0 -> 77.93 | Improved |
| add HumanSpawn() to spawn more human instances during curation (ablation) | SkyScenes · UAVid (SkyScenes->UAVid) | human mIoU: 4.71 -> 10.21 | Improved |
| add HumanSpawn() to spawn more human instances during curation (ablation) | SkyScenes · SkyScenes in-distribution | human mIoU: 43.03 -> 61.79 | Improved |
| add SkyScenes data jointly with 5% real UAVid labels | DeepLabv2 (R-101) · UAVid 5% labeled | mIoU: 48.25 -> 59.27 | Improved |
| add SkyScenes joint training or finetuning with full real data | DAFormer and Rein · UAVid 100% labeled | mIoU: 72.47 -> 72.25 (JT); 76.55 -> 76.54 (JT) | No clear effect |
| add depth modality to RGB | M3L multimodal segmentation · SkyScenes test | average IoU: 80.01 -> 83.22 | Improved |
| train on one weather and test on another (Clear to Rainy) | DAFormer · SkyScenes weather split | mIoU: 73.91 -> 69.95 | Hurt |
| train on Noon and test on Night | DAFormer · SkyScenes daytime split | mIoU: 73.91 -> 35.60 | Hurt |
| train on Rural and test on Urban | DAFormer · SkyScenes map split | mIoU: 58.00 -> 35.90 | Hurt |
Closest papers by meaning
- AerialMegaDepth: Learning Aerial-Ground Reconstruction and View Synthesis (CVPR 2025 · similarity 0.87)
- Aerial Lifting: Neural Urban Semantic and Building Instance Lifting from Aerial Imagery (CVPR 2024 · similarity 0.86)
- WeatherPrompt: Multi-modality Representation Learning for All-Weather Drone Visual Geo-Localization (NeurIPS 2025 · similarity 0.85)
- SkyMoE: A Vision-Language Foundation Model for Enhancing Geospatial Interpretation with Mixture of Experts (AAAI 2026 · similarity 0.85)
- PixDLM: A Dual-Path Multimodal Language Model for UAV Reasoning Segmentation (CVPR 2026 · similarity 0.85)
- SkySense-O: Towards Open-World Remote Sensing Interpretation with Vision-Centric Visual-Language Modeling (CVPR 2025 · similarity 0.85)
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.
