Research map / Point clouds / Segmentation and detection
ProposalContrast: Unsupervised Pre-training for LiDAR-Based 3D Object Detection
Junbo Yin, Dingfu Zhou, Liangjun Zhang, Jin Fang, Cheng-Zhong Xu, Jianbing Shen, Wenguan 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 attentive proposal encoder instead of max pooling (ablation) | ProposalContrast, CenterPoint · Waymo val 20% | mAPH: 59.32 -> 60.13 (+0.81 mAPH) | Improved |
| pre-train with proposal-level contrast instead of random init (ablation) | CenterPoint VoxelNet · Waymo val 20% | mAPH: 56.96 -> 60.13 (+3.17) | Improved |
| pre-train at proposal level rather than point level (PointContrast) (ablation) | CenterPoint VoxelNet · Waymo val 20% | mAPH: 57.75 -> 60.13 | Improved |
| pre-train at proposal level rather than scene level (DepthContrast) (ablation) | CenterPoint VoxelNet · Waymo val 20% | mAPH: 57.96 -> 60.13 | Improved |
| add inter-cluster separation (ICS) to inter-proposal discrimination (IPD) (ablation) | ProposalContrast · Waymo val 20% | mAPH: 59.56 -> 60.13 | Improved |
| add Sinkhorn-Knopp clustering to ICS (ablation) | ICS-only pre-training · Waymo val 20% | mAPH: 54.52 -> 59.16 (-2.44 mAPH without SKC relative to baseline) | Improved |
| use 2048 proposals instead of 1024 (ablation) | ProposalContrast · Waymo val 20% | mAPH: 59.47 -> 60.13 | Improved |
| use 2048 proposals instead of 4096 (ablation) | ProposalContrast · Waymo val 20% | mAPH: 59.91 -> 60.13 | Improved |
| use radius 1.0 m instead of 0.5 m or 2.0 m (ablation) | ProposalContrast · Waymo val 20% | mAPH: 59.81 (0.5) / 59.55 (2.0) -> 60.13 (1.0) | Improved |
| use 128 clusters instead of 64 or 256 in ICS (ablation) | ICS-only pre-training · Waymo val 20% | mAPH: 58.87 (64), 59.16 (128), 58.91 (256) | Mixed |
| pre-train with ProposalContrast on small labeled fraction (1%) | CenterPoint VoxelNet · Waymo val 1% labels | APH: 17.83 -> 34.78 | Improved |
| pre-train with ProposalContrast when labels are plentiful (100%) | PointPillars / VoxelNet · Waymo val 100% labels | APH: +0.40 / +0.61 APH | No clear effect |
Later papers that compared with this method
3 later accepted papers report ProposalContrast as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Self-Supervised Learning of LiDAR 3D Point Clouds via 2D-3D Neural Calibration (TPAMI 2025 · better on 1 of 1 shared result)
- Pre-training LiDAR-based 3D Object Detectors through Colorization (ICLR 2024 · better on 1 of 2 shared results)
- PRED: Pre-training via Semantic Rendering on LiDAR Point Clouds (NeurIPS 2023 · better on 2 of 2 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
- PTNET: A PROPOSAL-CENTRIC TRANSFORMER NET- WORK FOR 3D OBJECT DETECTION (ICLR 2026 · similarity 0.92)
- PointRCNN: 3D Object Proposal Generation and Detection From Point Cloud (CVPR 2019 · similarity 0.90)
- Not All Points Are Equal: Learning Highly Efficient Point-Based Detectors for 3D LiDAR Point Clouds (CVPR 2022 · similarity 0.90)
- Towards Unsupervised Object Detection From LiDAR Point Clouds (CVPR 2023 · similarity 0.90)
- VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (CVPR 2018 · similarity 0.90)
- STONE: A Submodular Optimization Framework for Active 3D Object Detection (NeurIPS 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.
