Research map / Detection and segmentation / 3D detection and driving
DSGN: Deep Stereo Geometry Network for 3D Object Detection
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 |
|---|---|---|---|
| Use DSGN instead of PL++: P-RCNN | KITTI Car 3D detection · KITTI test, Car moderate | 3D AP: 42.43 -> 52.18 | Helped |
| Use DSGN instead of PL++: P-RCNN | KITTI Car 3D detection · KITTI val, Car moderate | 3D AP: 50.1 -> 54.27 | Helped |
| Add 3D-volume point-cloud supervision to monocular IMG→3DV (ablation) | DSGN · KITTI val, Car moderate | 3D AP: 6.22 -> 13.66 | Helped |
| Add 3D-volume point-cloud supervision to stereo IMG→3DV (ablation) | DSGN · KITTI val, Car moderate | 3D AP: 11.03 -> 42.57 | Helped |
| Use plane-sweep rather than disparity-based cost volume (ablation) | Stereo cost-volume-to-3D pipeline · KITTI val, Car moderate | 3D AP: 45.89 -> 54.27 | Helped |
| Supervise the plane-sweep cost volume (ablation) | Stereo IMG→PSCV→3DV · KITTI val, Car moderate | 3D AP: 38.48 -> 54.27 | Helped |
| Use stereo rather than monocular input with 3D-volume supervision (ablation) | IMG→3DV · KITTI val, Car moderate | 3D AP: 13.66 -> 42.57 | Helped |
| Use last cost-volume features rather than binary occupancy (ablation) | 3DGV representation · KITTI val, Car moderate | 3D AP: 37.86 -> 54.27 | Helped |
| Use last cost-volume features rather than occupancy probabilities (ablation) | 3DGV representation · KITTI val, Car moderate | 3D AP: 43.24 -> 54.27 | Helped |
| Jointly optimize box corners instead of optimizing box parameters separately (ablation) | DSGN box regression · KITTI val, Car moderate | 3D AP: 40.71 -> 45.51 | Helped |
| Add mapped left-image features directly to 3DGV (ablation) | DSGN with joint box regression · KITTI val, Car moderate | 3D AP: 45.51 -> 44.79 | Hurt |
| Weight mapped image features by depth probability (ablation) | DSGN with joint box regression · KITTI val, Car moderate | 3D AP: 45.51 -> 46.52 | Helped |
Closest papers by meaning
- Triangulation Learning Network: From Monocular to Stereo 3D Object Detection (CVPR 2019 · similarity 0.89)
- 3D Bounding Box Estimation Using Deep Learning and Geometry (CVPR 2017 · similarity 0.88)
- Stereo R-CNN Based 3D Object Detection for Autonomous Driving (CVPR 2019 · similarity 0.88)
- Pseudo-LiDAR From Visual Depth Estimation: Bridging the Gap in 3D Object Detection for Autonomous Driving (CVPR 2019 · similarity 0.88)
- Disp R-CNN: Stereo 3D Object Detection via Shape Prior Guided Instance Disparity Estimation (CVPR 2020 · similarity 0.88)
- Uni3D: A Unified Baseline for Multi-Dataset 3D Object Detection (CVPR 2023 · similarity 0.88)
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.