Research map / Point clouds / Segmentation and detection
DetZero: Rethinking Offboard 3D Object Detection with Long-term Sequential Point Clouds
Tao Ma, Xuemeng Yang, Hongbin Zhou, Xin Li, Botian Shi, Junjie Liu, Yuchen Yang, Zhizheng Liu, Liang He, Yu Qiao, Yikang Li, 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 |
|---|---|---|---|
| add attribute-based refining modules (geometry, position, confidence) on top of upstream tracks (ablation) | DetZero · WOD val Vehicle | APH L2: 75.24 -> 82.92 | Improved |
| add geometry refining model (ablation) | DetZero upstream · WOD val Vehicle IoU=0.7 | 3D APH L1: 82.57 -> 84.49 | Improved |
| add position refining model (ablation) | DetZero upstream · WOD val Vehicle IoU=0.7 | 3D APH L1: 82.57 -> 85.48 | Improved |
| add confidence refining model (ablation) | DetZero GRM+PRM · WOD val Vehicle IoU=0.7 | 3D APH L1: 87.81 -> 89.06 | Improved |
| add offline tracker over baseline tracker in the detection pipeline (ablation) | DetZero upstream · WOD val Vehicle IoU=0.7 | 3D APH L1: 82.57 -> 82.57 | No clear effect |
| use own offline tracker instead of baseline tracker (ablation) | DetZero · WOD val Vehicle | Recall@track: 23.96 -> 40.28 | Improved |
| replace SOTA trackers with the proposed offline tracker (ablation) | DetZero · WOD val Vehicle | 3D APH: 80.56 -> 82.92 (vs ImmortalTracker) | Improved |
| use attribute-based refiner instead of 3DAL refiner (ablation) | DetZero with baseline tracker · WOD val Vehicle | 3D APH L2: 76.33 -> 78.14 | Improved |
| train student on 10% human + 90% DetZero auto labels versus 10% human only | CenterPoint student · WOD Vehicle | APH: 66.28 -> 73.91 | Improved |
| train student on 90% DetZero labels only versus 100% human labels | CenterPoint student · WOD Pedestrian | APH: 71.16 -> 71.39 | Improved |
| apply refining module to lower quality upstream detections (ablation) | DetZero refiner · WOD val Vehicle | APH L2: 70.00 -> 74.68 | 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.
- CenterFormer — CenterFormer: Center-based Transformer for 3D Object Detection (ECCV 2022) · better on 1 of 1 shared result
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
- V-DETR: DETR with Vertex Relative Position Encoding for 3D Object Detection (ICLR 2024 · similarity 0.87)
- MAD: Memory-Augmented Detection of 3D Objects (CVPR 2025 · similarity 0.86)
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.86)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.86)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.86)
- AS-Det: Active Sampling for Adaptive 3D Object Detection in Point Clouds (AAAI 2025 · similarity 0.86)
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.
