Research map / 3D vision / Depth and stereo
GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose
Read the paperCheck 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 |
|---|---|---|---|
| Use GeoNet rather than Zhou et al. updated | Monocular depth estimation · KITTI 2015 Eigen split | Abs Rel: 0.183 -> 0.155 | Helped |
| Use GeoNet rather than Godard et al. | Monocular depth estimation · KITTI 2015 Eigen split | Abs Rel: 0.148 -> 0.155 | Hurt |
| Use GeoNet rather than FlowNet2 | Optical flow estimation · KITTI 2015 flow training set | EPE All: 10.06 -> 10.81 | Hurt |
| Add geometric consistency loss to direct-flow FlowNetS (ablation) | DirFlowNetS · KITTI 2015 flow training set | EPE All: 12.86 -> 12.21 | Helped |
| Use residual flow learning and the GeoNet architecture instead of direct flow learning (ablation) | Optical flow estimation · KITTI 2015 flow training set | EPE All: 12.21 -> 10.81 | Helped |
| Replace adaptive consistency weighting with naive unweighted consistency loss (ablation) | GeoNet · KITTI 2015 flow training set | EPE All: 10.81 -> 17.18 | Hurt |
| Use ResNet50 instead of VGG (ablation) | GeoNet monocular depth estimator · KITTI 2015 Eigen split | Abs Rel: 0.164 -> 0.155 | Helped |
| Finetune the full network after stage-wise training (ablation) | GeoNet · KITTI | performance | No clear effect |
Closest papers by meaning
- Feature-Level Collaboration: Joint Unsupervised Learning of Optical Flow, Stereo Depth and Camera Motion (CVPR 2021 · similarity 0.92)
- Unsupervised Learning of Depth and Ego-Motion From Monocular Video Using 3D Geometric Constraints (CVPR 2018 · similarity 0.91)
- Unsupervised Learning of Depth and Ego-Motion From Video (CVPR 2017 · similarity 0.91)
- Flow-NeRF: Joint Learning of Geometry, Poses, and Dense Flow within Unified Neural Representations (CVPR 2025 · similarity 0.89)
- E-MoFlow: Learning Egomotion and Optical Flow from Event Data via Implicit Regularization (NeurIPS 2025 · similarity 0.88)
- Zero-Shot Monocular Scene Flow Estimation in the Wild (CVPR 2025 · similarity 0.87)
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.