Research map / Detection and segmentation / 3D detection and driving
Fully Sparse 3D Object Detection
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 |
|---|---|---|---|
| Use the fully sparse FSD detector instead of PV-RCNN++ (center) | Waymo Open Dataset detection · Waymo Open Dataset validation | mAP/mAPH L2: 71.7/69.5 -> 72.9/70.8 | Helped |
| Use FSD instead of re-implemented CenterPoint | Argoverse 2 detection · Argoverse 2 validation | Average precision: 22.0 -> 24.0 | Helped |
| Use FSD instead of re-implemented CenterPoint | Argoverse 2 detection · Argoverse 2 validation | Composite Score: 17.6 -> 19.1 | Helped |
| Replace dense CenterPoint with fully sparse FSD | Argoverse 2 inference · Argoverse 2 validation, full area | Latency (ms): 232 -> 97 | Helped |
| Add point grouping with pooled group features (ablation) | FSD plain · Waymo Open Dataset validation | L2 3D APH Vehicle: 62.29 -> 63.13 | Helped |
| Add SIR point feature extraction after grouping (ablation) | FSD agg · Waymo Open Dataset validation | L2 3D APH Vehicle: 63.13 -> 65.20 | Helped |
| Add group correction and SIR2 (ablation) | FSD nogc · Waymo Open Dataset validation | L2 3D APH Vehicle: 65.20 -> 69.30 | Helped |
| Avoid point downsampling in SIR; use raw points rather than 30cm voxels (ablation) | FSD SIR · Argoverse 2 validation | AP Construction Cone: 36.5 -> 38.6 | Helped |
| Restrict detection to the map region of interest (ablation) | FSD · Argoverse 2 validation, RoI only | mAP: 24.0 -> 23.2 | Hurt |
| Remove the ground region (ablation) | FSD · Argoverse 2 validation, without ground | mAP: 24.0 -> 21.0 | Hurt |
| Use FSD nogc in place of the feature-diffusion SST center head (ablation) | Waymo vehicle detection · Waymo Open Dataset validation, vehicles length [12, +infty) | AP: 30.5 -> 47.9 | Helped |
| Add dense feature diffusion and center-based assignment (ablation) | SST center head · Waymo vehicle detection, vehicles length [12, +infty) | AP: 42.2 -> 30.5 | Hurt |
Closest papers by meaning
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.92)
- Fully Convolutional One-Stage 3D Object Detection on LiDAR Range Images (NeurIPS 2022 · similarity 0.91)
- LSK3DNet: Towards Effective and Efficient 3D Perception with Large Sparse Kernels (CVPR 2024 · similarity 0.90)
- Far3D: Expanding the Horizon for Surround-View 3D Object Detection (AAAI 2024 · similarity 0.90)
- SwiftPillars: High-Efficiency Pillar Encoder for Lidar-Based 3D Detection (AAAI 2024 · similarity 0.89)
- 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.