Research map / 3D vision / Depth and stereo
MegaDepth: Learning Single-View Depth Prediction From Internet Photos
Zhengqi Li, Noah Snavely
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 multi-scale gradient matching loss (ablation) | MegaDepth depth predictor · MegaDepth test | si-RMSE: 0.148 -> 0.123 | Improved |
| Add multi-scale gradient matching loss (ablation) | MegaDepth depth predictor · MegaDepth test | SDR %: 31.75 -> 27.11 | Improved |
| Add automatically labeled ordinal depth loss (ablation) | MegaDepth depth predictor with L_data + L_grad · MegaDepth test | si-RMSE: 0.123 -> 0.104 | Improved |
| Add automatically labeled ordinal depth loss (ablation) | MegaDepth depth predictor with L_data + L_grad · MegaDepth test | SDR %: 27.11 -> 25.82 | Improved |
| Add automatically labeled ordinal depth loss (ablation) | MegaDepth depth predictor with L_data + L_grad · MegaDepth test | SDR=: 26.17 -> 27.73 | Hurt |
| Use refined and cleaned MVS depth maps instead of raw MVS depth maps (ablation) | MegaDepth training data · Make3D | RMS: 11.41 -> 5.322 | Improved |
| Use refined and cleaned MVS depth maps instead of raw MVS depth maps (ablation) | MegaDepth training data · KITTI | RMS: 12.15 -> 6.680 | Improved |
| Use refined and cleaned MVS depth maps instead of raw MVS depth maps (ablation) | MegaDepth training data · DIW | WHDR %: 31.32 -> 24.55 | Improved |
| Use an hourglass architecture instead of VGG | MegaDepth depth predictor · MegaDepth test | si-RMSE: 0.115 -> 0.104 | Improved |
| Train on MegaDepth instead of Make3D | Single-view depth predictor · Make3D | RMS: 4.45 -> 6.23 | Mixed |
| Train on MegaDepth instead of KITTI | Single-view depth predictor · KITTI | RMS: 5.93 -> 6.68 | Mixed |
| Train on MegaDepth instead of DIW | Single-view depth predictor · DIW | WHDR %: 22.14 -> 24.55 | Hurt |
Later papers that compared with this method
3 later accepted papers report MegaDepth as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Physics-informed Knowledge Transfer for Underwater Monocular Depth Estimation (ECCV 2024 · better on 1 of 1 shared result)
- Outdoor inverse rendering from a single image using multiview self-supervision (TPAMI 2021 · better on 2 of 2 shared results)
- Learning Single-Image Depth From Videos Using Quality Assessment Networks (CVPR 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
- MegaSynth: Scaling Up 3D Scene Reconstruction with Synthesized Data (CVPR 2025 · similarity 0.88)
- MVSAnywhere: Zero-Shot Multi-View Stereo (CVPR 2025 · similarity 0.87)
- UniDepth: Universal Monocular Metric Depth Estimation (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)
- DCDepth: Progressive Monocular Depth Estimation in Discrete Cosine Domain (NeurIPS 2024 · similarity 0.87)
- SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation (AAAI 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.
