Research map / Detection and segmentation / 3D detection and driving
SparseLIF: High-Performance Sparse LiDAR-Camera Fusion for 3D Object Detection
Hongcheng Zhang, Liu Liang, Pengxin Zeng, Xiao Song, Zhe Wang
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 |
|---|---|---|---|
| add Perspective-Aware Query Generation (PAQG) to RIAS (ablation) | SparseLIF-S · nuScenes val | NDS: 73.3 -> 74.3 | Improved |
| add Uncertainty-Aware Fusion (UAF) to RIAS (ablation) | SparseLIF-S · nuScenes val | NDS: 73.3 -> 74.1 | Improved |
| add RoI-Aware Sampling (RIAS) (ablation) | SparseLIF-S with PAQG and UAF · nuScenes val | NDS: 70.8 -> 74.6 | Improved |
| add PAQG for distant objects (beyond 30m) (ablation) | SparseLIF-S · nuScenes val, objects beyond 30m | mAP: 56.9 -> 58.5 | Improved |
| add PAQG for small objects (barrier) (ablation) | SparseLIF-S · nuScenes val, barrier 20-30m | AP: 54.5 -> 63.1 | Improved |
| add UAF under limited LiDAR FOV of 120 degrees (ablation) | SparseLIF-T · nuScenes val, LiDAR FOV 120 degrees | NDS: 62.1 -> 65.2 | Improved |
| add UAF under camera occlusion and unsynchronized sensors (ablation) | SparseLIF-T · nuScenes val, camera occlusion and stuck | NDS: 71.4 -> 72.0; 75.9 -> 76.5 | Improved |
| use temporal multi-frame input (13 frames) instead of single frame | SparseLIF · nuScenes val | NDS: 74.6 -> 77.5 | Improved |
| add simple self-model ensemble without TTA | SparseLIF-T · nuScenes test | NDS: 77.0 -> 77.7 | Improved |
Closest papers by meaning
- SparseFusion: Fusing Multi-Modal Sparse Representations for Multi-Sensor 3D Object Detection (ICCV 2023 · similarity 0.93)
- Fully Sparse Fusion for 3D Object Detection (TPAMI 2024 · similarity 0.93)
- Fully Sparse 3D Object Detection (NeurIPS 2022 · similarity 0.91)
- Sparse Fuse Dense: Towards High Quality 3D Detection With Depth Completion (CVPR 2022 · similarity 0.91)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.90)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 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.
