Research map / Detection and segmentation / 3D detection and driving
Object DGCNN: 3D Object Detection using Dynamic Graphs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use Object DGCNN with a SparseConv backbone instead of CenterPoint with the same backbone | 3D object detection on nuScenes · nuScenes | NDS: 64.19 -> 66.10 | Helped |
| Use Object DGCNN with a PointPillars backbone instead of CenterPoint with the same backbone | 3D object detection on nuScenes · nuScenes | mAP: 47.48 -> 53.31 | Helped |
| Remove NMS from CenterPoint inference | CenterPoint (voxel) · nuScenes | NDS: 64.19 -> 57.00 | Hurt |
| Add set-to-set distillation from a SparseConv teacher to a PointPillars student | PointPillars · nuScenes | NDS: 62.80 -> 63.37 | Helped |
| Replace feature-map distillation with set-to-set distillation | PointPillars student with SparseConv teacher · nuScenes | mAP: 53.29 -> 53.89 | Helped |
| Add self-distillation | PointPillars · nuScenes | NDS: 62.80 -> 63.41 | Helped |
| Add self-distillation | SparseConv detector · nuScenes | NDS: 66.04 -> 66.45 | Helped |
| Use dense point clouds for the teacher while the student receives sparse point clouds | PointPillars student · nuScenes | NDS: 42.12 -> 42.79 | Helped |
| Replace DGCNN with multi-head self-attention for object interactions (ablation) | Object DGCNN · nuScenes ablation | NDS: 41.32 -> 39.89 | Hurt |
| Use 6 DGCNN layers instead of 1 (ablation) | Object DGCNN · nuScenes ablation | NDS: 35.91 -> 41.32 | Helped |
| Use 16 nearest neighbors instead of 1 (ablation) | DGCNN object interactions · nuScenes ablation | NDS: 40.21 -> 41.32 | Helped |
| Increase neighbors from 16 to 64 (ablation) | DGCNN object interactions · nuScenes ablation | NDS: 41.32 -> 39.80 | Hurt |
Closest papers by meaning
- Towards Efficient 3D Object Detection with Knowledge Distillation (NeurIPS 2022 · similarity 0.89)
- PointDistiller: Structured Knowledge Distillation Towards Efficient and Compact 3D Detection (CVPR 2023 · similarity 0.88)
- OD$^3$: Optimization-free Dataset Distillation for Object Detection (ICLR 2026 · similarity 0.88)
- A Hierarchical Graph Network for 3D Object Detection on Point Clouds (CVPR 2020 · similarity 0.88)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.87)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · 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.