Research map / Point clouds / Segmentation and detection
From Points to Parts: 3D Object Detection from Point Cloud with Part-aware and Part-aggregation Network
Shaoshuai Shi, Zhe Wang, Jianping Shi, Xiaogang Wang, Hongsheng Li
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 SparseConvUNet backbone instead of PointNet++ (ablation) | Part-A2-free · KITTI val car | recall IoU 0.7 (100 proposals): 74.81 -> 81.54 | Improved |
| use RoI-aware pooling instead of fixed-sized RoI pooling (ablation) | Part-A2-anchor · KITTI val car | 3D AP moderate: 78.61 -> 79.47 | Improved |
| use sparse conv instead of FC layers in stage-II (ablation) | Part-A2-anchor · KITTI val car | 3D AP moderate: 79.32 -> 79.47 | No clear effect |
| increase RoI pooling size from 6 to 14 (ablation) | Part-A2-anchor · KITTI val car | 3D AP moderate: 78.85 -> 79.47 | Improved |
| use pooling size 14 instead of 7 with FC layers (ablation) | Part-A2-anchor · KITTI val car | 3D AP moderate: 79.11 -> 79.32 | Improved |
| use sparse conv instead of FC at pooling size 7 (ablation) | Part-A2-anchor · KITTI val car | 3D AP moderate: 79.11 -> 79.21 | Improved |
| use anchor-based proposals instead of anchor-free (ablation) | Part-A2 · KITTI val car | recall IoU 0.7 (50 proposals): 77.12 -> 83.71 | Improved |
| anchor-free vs anchor-based proposals (ablation) | Part-A2 · KITTI val cyclist and pedestrian | 3D AP: 73.35 vs 73.07 | No clear effect |
| use anchor-based instead of anchor-free proposals for cars (ablation) | Part-A2 · KITTI val car | 3D AP moderate: 78.96 -> 79.47 | Improved |
| use bin-based center loss with anchor-free proposals (ablation) | Part-A2-free · KITTI val car | recall convergence | Improved |
| use residual-based center loss with anchor-based proposals (ablation) | Part-A2-anchor · KITTI val car | recall convergence | Improved |
| add intra-object part location supervision (ablation) | Part-A2-anchor · KITTI val car | recall and AP | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PointRCNN — PointRCNN: 3D Object Proposal Generation and Detection From Point Cloud (CVPR 2019) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- AGO-Net: Association-Guided 3D Point Cloud Object Detection Network (TPAMI 2021 · similarity 0.92)
- PillarNeXt: Rethinking Network Designs for 3D Object Detection in LiDAR Point Clouds (CVPR 2023 · similarity 0.91)
- Not All Points Are Equal: Learning Highly Efficient Point-Based Detectors for 3D LiDAR Point Clouds (CVPR 2022 · similarity 0.91)
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.90)
- PointRCNN: 3D Object Proposal Generation and Detection From Point Cloud (CVPR 2019 · similarity 0.90)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · 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.
