Research map / 3D vision / Depth and stereo
LEGO: Learning Edge With Geometry All at Once by Watching Videos
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 |
|---|---|---|---|
| Add the joint 3D-ASAP depth, normal, and edge regularization | LEGO · KITTI 2015 KITTI split | Abs Rel: 0.165 -> 0.154 | Helped |
| Add the joint 3D-ASAP depth, normal, and edge regularization | LEGO · KITTI 2015 Eigen split | Abs Rel: 0.182 -> 0.162 | Helped |
| Train with the joint 3D-ASAP edge and geometry approach | LEGO · Cityscapes to KITTI | Abs Rel: 0.275 -> 0.201 | Helped |
| Train with the joint 3D-ASAP edge and geometry approach | LEGO · Cityscapes to Make3D | RMSE: 10.47 -> 7.194 | Helped |
| Add joint depth and normal edge constraints rather than using only the strongest single edge source (ablation) | LEGO · Cityscapes | ODS: 0.679 -> 0.710 | Helped |
| Remove the fly-out mask (ablation) | LEGO · KITTI 2015 KITTI split | Abs Rel: 0.154 -> 0.157 | Hurt |
| Remove the fly-out mask (ablation) | LEGO · KITTI 2015 Eigen split | Abs Rel: 0.162 -> 0.170 | Hurt |
| Apply geometrical edges only in the depth smoothness term (ablation) | LEGO · Cityscapes | ODS: 0.602 -> 0.710 | Hurt |
| Apply geometrical edges only in the normal smoothness term (ablation) | LEGO · Cityscapes | ODS: 0.679 -> 0.710 | Hurt |
| Apply geometrical edges only in the depth smoothness term (ablation) | LEGO · Cityscapes | AP: 0.594 -> 0.729 | Hurt |
| Apply geometrical edges only in the normal smoothness term (ablation) | LEGO · Cityscapes | AP: 0.692 -> 0.729 | Hurt |
Closest papers by meaning
- Unsupervised Learning of Geometry From Videos With Edge-Aware Depth-Normal Consistency (AAAI 2018 · similarity 0.90)
- Geometry-Aware Network for Non-Rigid Shape Prediction From a Single View (CVPR 2018 · similarity 0.88)
- Joint Graph-Based Depth Refinement and Normal Estimation (CVPR 2020 · similarity 0.87)
- Learning 3D Object Shape and Layout Without 3D Supervision (CVPR 2022 · similarity 0.87)
- P3Depth: Monocular Depth Estimation With a Piecewise Planarity Prior (CVPR 2022 · similarity 0.87)
- Edge-Aware 3D Instance Segmentation Network with Intelligent Semantic Prior (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.