Research map / 3D vision / Depth and stereo
Single-Stage 3D Geometry-Preserving Depth Estimation Model Training on Dataset Mixtures With Uncalibrated Stereo Data
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SVDE models with GP² on UTS and UTSS data rather than UTS data only | B5-LRN on DIMLI+RW+MD+MV · TUM | δ1.25: 0.145 -> 0.130 | Helped |
| Train with GP² on the LeReS mixture | B5-LRN · NYU | δ1.25: 0.106 -> 0.095 | Helped |
| Train with GP² on the LeReS mixture | B5-LRN · Sintel | rel: 0.397 -> 0.320 | Helped |
| Train with GP² on the LeReS mixture | B5-LRN · ETH3D | rel: 0.174 -> 0.131 | Helped |
| Train with GP² on the LeReS mixture | B5-LRN · iBims-1 | rel: 0.100 -> 0.092 | Helped |
| Replace LeReS+PCM training with GP² training | LeReS on DIMLI+RW+MD+MV · TUM | δ1.25: 0.193 -> 0.139 | Helped |
| Use GP²-trained B5-LRN depth estimates with the same focal-length PCM | B5-LRN+PCM versus LeReS+PCM · 2D-3D-S | Point cloud RMSE: 0.429 -> 0.363 | Helped |
| Use B5-LRN without PCM rather than LeReS+PCM | General-purpose depth estimation inference · Inference | Inference time: 201.9 -> 35.4 | Helped |
| Add UTSS data to UTS training data using GP² (ablation) | B5-LRN trained on NYU Raw · NYU Raw test | Depth-estimation accuracy | No clear effect |
| Use only 10-20% UTS data alongside UTSS data (ablation) | B5-LRN trained with GP² · NYU Raw test | Depth-estimation accuracy | No clear effect |
Closest papers by meaning
- Learning To Recover 3D Scene Shape From a Single Image (CVPR 2021 · similarity 0.88)
- MVD-Fusion: Single-view 3D via Depth-consistent Multi-view Generation (CVPR 2024 · similarity 0.87)
- GeoDepth: From Point-to-Depth to Plane-to-Depth Modeling for Self-Supervised Monocular Depth Estimation (CVPR 2025 · similarity 0.87)
- Leveraging Global Stereo Consistency for Category-Level Shape and 6D Pose Estimation from Stereo Images (CVPR 2025 · similarity 0.86)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.86)
- UniDepth: Universal Monocular Metric Depth Estimation (CVPR 2024 · similarity 0.86)
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.