Research map / 3D vision / Depth and stereo
Sparse Beats Dense: Rethinking Supervision in Radar-Camera Depth Completion
Huadong Li, Minhao Jing, Jin Wang, Shichao Dong, Jiajun Liang, Haoqiang Fan, Renhe Ji
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 |
|---|---|---|---|
| Use single-frame sparse LiDAR supervision with Disruption-Compensation instead of multi-frame dense supervision | radar-camera depth completion · nuScenes (0-80m) | MAE (mm): 2179.3 -> 1927.0 (vs RadarNet) | Improved |
| Add Radar-Position Injection Module (ablation) | Disruption-Compensation framework · nuScenes (0-80m) | MAE (mm): 2201.1 -> 2116.4 | Improved |
| Add Radar-aware Mask Decoder (training only) (ablation) | Disruption-Compensation framework · nuScenes (0-80m) | MAE (mm): 2116.4 -> 1927.0 | Improved |
| Give dense (multi-frame or interpolated) supervision to the same backbone (ablation) | RadarNet, DORN, RC-PDA with ResNeXt backbone · nuScenes (0-80m) | MAE (mm): 1927.0 -> 2820.1 to 3104.5 | Hurt |
| Retrain LiDAR-camera completion methods on radar input without modification (ablation) | CompletionFormer, DySPN · nuScenes (0-80m) | MAE (mm): 1927.0 -> 3100.7 and 3549.2 | Hurt |
| Add Camera Intrinsics Disruption (random upsample and crop) (ablation) | sparse-supervised model · nuScenes | OMAE | Improved |
| Add Radar Disruption (extend radar points into vertical lines) (ablation) | sparse-supervised model · nuScenes | OMAE | Improved |
| Add noise to Y/Z instead of extending radar points to vertical lines (ablation) | Radar Disruption · nuScenes | OMAE | Hurt |
Closest papers by meaning
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.89)
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.87)
- Mind The Edge: Refining Depth Edges in Sparsely-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.87)
- Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning (Student Abstract) (AAAI 2025 · similarity 0.86)
- TacoDepth: Towards Efficient Radar-Camera Depth Estimation with One-stage Fusion (CVPR 2025 · similarity 0.86)
- Masked Spatial Propagation Network for Sparsity-Adaptive Depth Refinement (CVPR 2024 · similarity 0.85)
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.
