Research map / Detection and segmentation / 3D detection and driving
Pseudo-LiDAR++: Accurate Depth for 3D Object Detection in Autonomous Driving
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 |
|---|---|---|---|
| Replace disparity loss with depth loss (ablation) | PSMNet for P-RCNN · KITTI val car, moderate, IoU=0.7 | AP_BE V: 56.0 -> 61.9 | Helped |
| Add a depth-grid cost volume to depth-loss training (ablation) | PSMNet for P-RCNN · KITTI val car, moderate, IoU=0.7 | AP_BE V: 61.9 -> 64.0 | Helped |
| Use depth loss and depth cost volume instead of disparity estimation (ablation) | PSMNet for P-RCNN · KITTI val car, hard, IoU=0.7 | AP_BE V: 52.7 -> 57.3 | Helped |
| Propagate landmark LiDAR depths across the graph rather than only replacing landmark depths (ablation) | SDN pseudo-LiDAR with P-RCNN · KITTI val car, moderate, IoU=0.7 | AP_BE V: 73.0 -> 76.9 | Helped |
| Use 4-beam LiDAR with GDC instead of stereo-only SDN (ablation) | SDN pseudo-LiDAR with P-RCNN · KITTI val car, hard, IoU=0.7 | AP_BE V: 57.3 -> 73.4 | Helped |
| Use graph-based correction instead of PnP (ablation) | SDN pseudo-LiDAR with P-RCNN · KITTI val car, hard, IoU=0.7 | AP_BE V: 67.2 -> 73.4 | Helped |
| Reduce GDC input from 4 LiDAR beams to 2 (ablation) | SDN pseudo-LiDAR with P-RCNN · KITTI val car, moderate, IoU=0.7 | AP_BE V: 76.9 -> 72.0 | Hurt |
| Replace stereo depth with landmark depths only, without propagation (ablation) | SDN pseudo-LiDAR with P-RCNN · KITTI val car, hard, IoU=0.7 | AP_BE V: 73.4 -> 67.4 | Hurt |
| Use the optimized rather than naive GDC implementation (ablation) | SDN + GDC with P-RCNN · KITTI val car, hard, IoU=0.7 | AP_BE V: 73.4 -> 73.1 | No clear effect |
| Add SDN depth cost volume after depth-loss training (ablation) | PSMNet stereo depth for AVOD · KITTI val car, moderate, IoU=0.7 | AP_BE V: 57.1 -> 63.7 | Helped |
| Add graph-based sparse LiDAR correction (ablation) | SDN with P-RCNN · KITTI val car, hard, IoU=0.7, depth 30-50 m | AP_BE V: 27.4 -> 49.9 | Helped |
Closest papers by meaning
- Pseudo-LiDAR From Visual Depth Estimation: Bridging the Gap in 3D Object Detection for Autonomous Driving (CVPR 2019 · similarity 0.96)
- Pseudo-Stereo for Monocular 3D Object Detection in Autonomous Driving (CVPR 2022 · similarity 0.93)
- End-to-End Pseudo-LiDAR for Image-Based 3D Object Detection (CVPR 2020 · similarity 0.92)
- LaserNet: An Efficient Probabilistic 3D Object Detector for Autonomous Driving (CVPR 2019 · similarity 0.90)
- IDA-3D: Instance-Depth-Aware 3D Object Detection From Stereo Vision for Autonomous Driving (CVPR 2020 · similarity 0.90)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.89)
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.