Research map / Point clouds / Segmentation and detection
Streaming Object Detection for 3-D Point Clouds
Wei Han, Zhengdong Zhang, Benjamin Caine, Brandon Yang, Christoph Sprunk, Ouais Alsharif, Jiquan Ngiam, Vijay Vasudevan, Jonathon Shlens, Zhifeng Chen
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 |
|---|---|---|---|
| restrict the detector to a single slice (localized receptive field) | Non-streaming baseline · Waymo vehicles, 16 slices | mAP: 51.0 -> 35.0 | Hurt |
| increase slices from 16 to 32 without state | Localized receptive field · Waymo vehicles | mAP: 35.0 -> 10.5 | Hurt |
| add global NMS over all slices (ablation) | Localized receptive field · Waymo vehicles, 32 slices | mAP: 10.5 -> 39.2 | Improved |
| add stateful NMS (k=1) | Localized receptive field · Waymo vehicles, 32 slices | mAP: 10.5 -> 39.0 | Improved |
| use stateful NMS instead of global NMS (ablation) | Global NMS · Waymo pedestrians, 16 slices | mAP: 54.8 -> 54.9 | No clear effect |
| add LSTM recurrence to stateful NMS | Streaming PointPillars · Waymo vehicles, 32 slices | mAP: 39.2 -> 48.9 | Improved |
| add LSTM recurrence to stateful NMS | Streaming detector · Waymo vehicles, objects subtending 0-5 deg | mAP: 9.0 -> 10.8 | Improved |
| add stateful NMS | Localized RF · Waymo vehicles, objects subtending >35 deg | mAP: 14.0 -> 68.0 | Improved |
| stream with 8 slices, stateful NMS and RNN, instead of full rotation | Non-streaming baseline · Waymo vehicles (PointPillars) | latency (ms): 116 -> 17 | Improved |
| scale up streaming model (larger grid) vs non-streaming baseline | PointPillars pedestrians · Waymo pedestrians | mAP: 54.9 -> 60.1 | Improved |
Closest papers by meaning
- Not All Points Are Equal: Learning Highly Efficient Point-Based Detectors for 3D LiDAR Point Clouds (CVPR 2022 · similarity 0.91)
- LaserNet: An Efficient Probabilistic 3D Object Detector for Autonomous Driving (CVPR 2019 · similarity 0.90)
- Towards Unsupervised Object Detection From LiDAR Point Clouds (CVPR 2023 · similarity 0.90)
- MGTANet: Encoding Sequential LiDAR Points Using Long Short-Term Motion-Guided Temporal Attention for 3D Object Detection (AAAI 2023 · similarity 0.90)
- Robust 3D Object Detection using Probabilistic Point Clouds from Single-Photon LiDARs (ICCV 2025 · similarity 0.89)
- Pre-training LiDAR-based 3D Object Detectors through Colorization (ICLR 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.
