Research map / Point clouds / Segmentation and detection
CT3D++: Improving 3D Object Detection with Keypoint-Induced Channel-wise Transformer
Xian-Sheng Hua, Sijia Cai, Na Zhao, Bing Deng, Qiao Liang, Min-Jian Zhao, Jieping Ye
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 extended channel-wise re-weighting decoder (ablation) | CT3D++ · Waymo 20% Vehicle L2 | AP: 65.2 -> 68.9 | Improved |
| add extended channel-wise re-weighting decoder (ablation) | CT3D++ · Waymo 20% Cyclist L2 | AP: 53.5 -> 61.1 | Improved |
| replace channel-wise re-weighting alone with standard cross-attention (ablation) | decoder · Waymo 20% Vehicle L2 | AP: 65.2 -> 65.3 | No clear effect |
| add geometric raw-point embedding to semantic-only embedding (ablation) | CT3D++ · Waymo 20% Vehicle L2 | AP: 64.1 -> 68.9 | Improved |
| add BEV semantic features to geometric-only embedding (ablation) | CT3D++ · Waymo 20% Cyclist L2 | AP: 60.5 -> 61.1 | Improved |
| add voxel features from RPN middle layers (ablation) | CT3D++ · Waymo 20% Vehicle L2 | AP: 68.9 -> 69.0 | No clear effect |
| use PBC instead of self-attention (ablation) | CT3D++ · Waymo 20% Vehicle L2 | AP: 64.5 -> 68.9 | Improved |
| use PBC instead of MLP (ablation) | CT3D++ · Waymo 20% Vehicle L2 | AP: 66.4 -> 68.9 | Improved |
| use PBC instead of self-attention at 32 epochs (ablation) | CT3D++ · Waymo 20% Vehicle L1 | AP: 72.9 -> 78.5 | Improved |
| decode on raw points instead of key points (ablation) | CT3D++ · Waymo 20% Vehicle L2 | AP: 68.6 -> 68.9 | Improved |
| use category-based instead of object-based raw point sampling (ablation) | CT3D++ · Waymo 20% Vehicle L2 | AP: 69.0 -> 68.9 | No clear effect |
| upgrade CT3D to CT3D++ with PointPillar RPN (ablation) | RPN refinement · Waymo 20% Vehicle L2 | AP: 62.6 -> 64.8 | 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 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
- Improving 3D Object Detection With Channel-Wise Transformer (ICCV 2021 · similarity 0.96)
- An End-to-End Transformer Model for 3D Object Detection (ICCV 2021 · similarity 0.90)
- 3DPPE: 3D Point Positional Encoding for Transformer-based Multi-Camera 3D Object Detection (ICCV 2023 · similarity 0.89)
- Uni3DETR: Unified 3D Detection Transformer (NeurIPS 2023 · similarity 0.89)
- Point2Seq: Detecting 3D Objects As Sequences (CVPR 2022 · similarity 0.88)
- A Simple Vision Transformer for Weakly Semi-supervised 3D Object Detection (ICCV 2023 · 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.
