Dimensions overall score 5.0

GitHub Code Pulse

No public code linked for this paper yet.

• TerraSky3D is a high-resolution large-scale 3D reconstruction dataset comprising 50,000 images divided into 150 ground, aerial, and mixed scenes.result 95%
• The dataset focuses on European landmarks and comes with curated calibration data, camera poses, and depth maps.result 90%
• Existing aerial imagery datasets like AerialMegaDepth rely on pseudo-synthetic images from 3D meshes, which introduces a significant domain gap compared to real photographs.limitation 85%
• The data collection pipeline uses COLMAP for structure-from-motion and APD-MVS for depth generation, with Mask2Former for semantic filtering.method 90%
• Cameras were pre-calibrated using a ChArUco board and OpenCV, achieving sub-pixel reprojection errors.technical 90%
• On the TerraSky3D test set for stereo pose estimation, the RoMa model outperforms other methods across ground, aerial, and mixed scenarios.result 85%
• The RoMa model's robustness is derived from a dense-warping paradigm, yielding up to a 10x increase in inlier counts compared to traditional point-to-point matching.technical 80%
• Images in TerraSky3D were collected in a reduced time span to ensure consistent lighting conditions and complete viewpoint coverage.method 85%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card