Research map / Point clouds / Point cloud classification
Ponder: Point Cloud Pre-training via Neural Rendering
Di Huang, Sida Peng, Tong He, Honghui Yang, Xiaowei Zhou, Wanli Ouyang
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 |
|---|---|---|---|
| pre-train point cloud encoder with neural rendering | VoteNet · ScanNet detection | AP50: 33.5 -> 41.0 | Improved |
| supervise with both depth and color instead of one (ablation) | Ponder · ScanNet detection | AP50: 40.9 -> 41.0 | Improved |
| use depth rendering target rather than color (ablation) | Ponder · ScanNet detection | AP50: 40.5 -> 40.9 | Improved |
| change masking ratio from 0% to 75% (ablation) | Ponder · ScanNet detection | AP50: 40.7 -> 41.7 | No clear effect |
| use multi-resolution feature volume 16+32+64 instead of 16 only (ablation) | Ponder · ScanNet detection | AP50: 40.7 -> 41.0 | No clear effect |
| use 5 RGB-D views instead of 1 (ablation) | Ponder · ScanNet detection | AP50: 40.1 -> 41.0 | Improved |
| add near-surface, free-space and Eikonal losses (ablation) | Ponder · ScanNet detection | AP50: 40.5 -> 41.0 | Improved |
| pre-train longer (20 to 100 epochs) (ablation) | Ponder · ScanNet detection | AP50: 38.7 -> 41.0 | Improved |
| initialize ConvONet encoder with IAE pre-training | ConvONet PointNet++ · SISD scene reconstruction | IoU: 77.8 -> 75.7 | Hurt |
| initialize ConvONet encoder with Ponder | ConvONet PointNet++ · SISD scene reconstruction | IoU: 77.8 -> 80.2 | Improved |
| use point color as extra input feature without pre-training (ablation) | VoteNet · ScanNet detection | AP50: 33.5 -> 33.4 | No clear effect |
Closest papers by meaning
- From Image Collections to Point Clouds With Self-Supervised Shape and Pose Networks (CVPR 2020 · similarity 0.89)
- GeoMAE: Masked Geometric Target Prediction for Self-Supervised Point Cloud Pre-Training (CVPR 2023 · similarity 0.89)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 2025 · similarity 0.88)
- A Unified Approach to Interpreting Self-supervised Pre-training Methods for 3D Point Clouds via Interactions (CVPR 2025 · similarity 0.88)
- Sparse Point Cloud Patches Rendering via Splitting 2D Gaussians (CVPR 2025 · similarity 0.88)
- Multi-View Representation is What You Need for Point-Cloud Pre-Training (ICLR 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.
