Research map / Point clouds / Segmentation and detection
Voxel Transformer for 3D Object Detection
Jiageng Mao, Yujing Xue, Minzhe Niu, Haoyue Bai, Jiashi Feng, Xiaodan Liang, Hang Xu, Chunjing Xu
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 |
|---|---|---|---|
| replace SECOND 3D conv backbone with VoTr | SECOND · Waymo val vehicle LEVEL_1 | 3D mAP: 67.94 -> 68.99 (+1.05%) | Improved |
| replace PV-RCNN 3D conv backbone with VoTr | PV-RCNN · Waymo val vehicle LEVEL_1 | 3D mAP: 74.06 -> 74.95 (+3.26%) | Improved |
| use VoTr backbone for far objects | VoTr-TSD · Waymo val vehicle 30-50m and 50m-Inf | 3D mAP: +3.37% (30-50m), +4.83% (50m-Inf) | Improved |
| replace SECOND backbone with VoTr | SECOND · KITTI val car moderate | AP_3D: 76.48 -> 78.27 (+1.79%) | Improved |
| replace PV-RCNN backbone with VoTr | PV-RCNN · KITTI val car moderate | AP_3D: 83.69 -> 84.04 (+0.35%) | Improved |
| replace PV-RCNN backbone with VoTr | PV-RCNN · KITTI test car easy | AP_3D: 90.25 -> 89.90 | Hurt |
| replace PV-RCNN backbone with VoTr | PV-RCNN · KITTI test car hard | AP_3D: 76.82 -> 79.14 | Improved |
| add Dilated Attention to Local Attention (ablation) | VoTr-SSD · KITTI val car moderate | AP_3D: 75.48 -> 78.27 (+2.79%) | Improved |
| add dropout with probability 0.3 (ablation) | VoTr-SSD · KITTI val car moderate | AP_3D: 78.27 -> 69.75 (-8.52%) | Hurt |
| add dropout with probability 0.1 (ablation) | VoTr-SSD · KITTI val car moderate | AP_3D: 78.27 -> 75.97 | Hurt |
| increase attending voxels from 24 to 48 (ablation) | VoTr-SSD · KITTI val car moderate | AP_3D: 77.08 -> 78.27 (+1.19%) | Improved |
| replace conv backbone with VoTr (parameters) | SECOND · KITTI | model parameters: 5.3M -> 4.8M | 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.
- PV-RCNN — PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020) · better on 6 of 6 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
- Voxel Set Transformer: A Set-to-Set Approach to 3D Object Detection From Point Clouds (CVPR 2022 · similarity 0.93)
- Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection (AAAI 2021 · similarity 0.92)
- HVPR: Hybrid Voxel-Point Representation for Single-Stage 3D Object Detection (CVPR 2021 · similarity 0.91)
- MsSVT++: Mixed-Scale Sparse Voxel Transformer With Center Voting for 3D Object Detection (TPAMI 2023 · similarity 0.91)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.89)
- Voxel Mamba: Group-Free State Space Models for Point Cloud based 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.
