Research map / Point clouds / Segmentation and detection
DiffRefine: Diffusion-based Proposal Specific Point Cloud Densification for Cross-Domain Object Detection
Sangyun Shin, Yuhang He, Xinyu Hou, Samuel Hodgson, Andrew Markham, Niki Trigoni
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 spatial context feature f_gen (ablation) | DiffRefine · Waymo->NuScenes (car) | APBEV/AP3D: 1.58/1.06 gain | Improved |
| add differentiable warping (gradient flow to refinement) (ablation) | DiffRefine · Waymo->NuScenes (car) | APBEV/AP3D: 0.59/0.43 gain | Improved |
| add both differentiable warping and f_gen (ablation) | DiffRefine · Waymo->NuScenes (car) | AP3D: 23.73 -> 25.25 | Improved |
| use diffusion instead of voxel classification for generation (ablation) | DiffRefine · Waymo->NuScenes (car) | APBEV/AP3D: 20.44 -> 25.25 AP3D at 32^3 | Improved |
| shrink voxel grid from 32 to 8 for diffusion (ablation) | DiffRefine · Waymo->NuScenes (car) | AP3D: 25.25 -> 23.74 | Hurt |
| shrink voxel grid from 32 to 16 for classification (ablation) | Voxel classification · Waymo->NuScenes (car) | AP3D: 20.44 -> 21.87 | Mixed |
| increase denoising steps from 2 to 6 (ablation) | DiffRefine · NuScenes->KITTI | AP3D: 0.12, 0.08, 0.06, 0.03 per step | Improved |
| DiffRefine over ST3D baseline with SECOND-IoU | SECOND-IoU · Waymo->KITTI (car) | APBEV/AP3D: 7.03/9.61 gain | Improved |
| DiffRefine over GroupExp-DA with SECOND-IoU | SECOND-IoU · NuScenes->KITTI (car) | APBEV/AP3D: 2.28/2.92 gain | Improved |
| DiffRefine over best method on Pedestrian and Cyclist | SECOND-IoU · NuScenes->KITTI | APBEV/AP3D: 1.97/1.43 Pedestrian; 1.35/0.81 Cyclist | Improved |
Closest papers by meaning
- DiffuBox: Refining 3D Object Detection with Point Diffusion (NeurIPS 2024 · similarity 0.92)
- MonoDiff: Monocular 3D Object Detection and Pose Estimation with Diffusion Models (CVPR 2024 · similarity 0.88)
- Generalized Diffusion Detector: Mining Robust Features from Diffusion Models for Domain-Generalized Detection (CVPR 2025 · similarity 0.87)
- One for All: Multi-Domain Joint Training for Point Cloud Based 3D Object Detection (NeurIPS 2024 · similarity 0.87)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.87)
- Adaptive 3D Reconstruction via Diffusion Priors and Forward Curvature-Matching Likelihood Updates (NeurIPS 2025 · similarity 0.87)
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.
