Research map / 3D vision / 3D generation
PonderV2: Improved 3D Representation With a Universal Pre-Training Paradigm
Haoyi Zhu, Honghui Yang, Xiaoyang Wu, Di Huang, Sha Zhang, Xianglong He, Hengshuang Zhao, Chunhua Shen, Yu Qiao, Tong He, 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 |
|---|---|---|---|
| add depth rendering supervision alongside color (ablation) | Ponder-RGBD VoteNet · ScanNet 3D detection | mAP@50: 40.5 -> 41.0 | Improved |
| add color rendering supervision alongside depth (ablation) | Ponder-RGBD VoteNet · ScanNet 3D detection | mAP@50: 40.9 -> 41.0 | Improved |
| use five input views instead of one (ablation) | Ponder-RGBD VoteNet · ScanNet 3D detection | mAP@50: 40.1 -> 41.0 | Improved |
| use five input views instead of one (ablation) | Ponder-RGBD VoteNet · SUN RGB-D 3D detection | mAP@50: 35.4 -> 36.6 | Improved |
| use 75% mask ratio instead of no masking (ablation) | Ponder-RGBD VoteNet · ScanNet 3D detection | mAP@50: 40.7 -> 41.7 | Improved |
| remove masking entirely (0% mask ratio) (ablation) | Ponder-RGBD VoteNet · SUN RGB-D 3D detection | mAP@50: 37.3 vs 37.0 at 75% | No clear effect |
| use multi-resolution feature volume 16+32+64 instead of 16 only (ablation) | Ponder-RGBD VoteNet · ScanNet 3D detection | mAP@50: 40.7 -> 41.0 | No clear effect |
| use mask ratio 0.3 for camera images instead of 0.1 (ablation) | PonderV2 UVTR-C · nuScenes | NDS: 31.9 -> 32.9 | Improved |
| increase decoder depth from (2,2) to (6,4) (ablation) | PonderV2 UVTR-C · nuScenes | NDS: 31.3 -> 32.9 | Improved |
| increase decoder width from 32 to 256 (ablation) | PonderV2 UVTR-C · nuScenes | NDS: 32.9 -> 32.4 | No clear effect |
| use NeuS rendering instead of UniSurf (ablation) | PonderV2 UVTR-C · nuScenes | NDS: 32.5 -> 32.9 | Improved |
| use NeuS rendering instead of VolSDF (ablation) | PonderV2 UVTR-C · nuScenes | NDS: 32.8 -> 32.9 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports Ponder-v2† as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- From Thousands to Billions: 3D Visual Language Grounding via Render-Supervised Distillation from 2D VLMs (ICML 2025 · better on 3 of 3 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
- Building a Strong Pre-Training Baseline for Universal 3D Large-Scale Perception (CVPR 2024 · similarity 0.88)
- Unifying 2D and 3D Vision-Language Understanding (ICML 2025 · similarity 0.87)
- VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction (CVPR 2026 · similarity 0.87)
- Uni3D: Exploring Unified 3D Representation at Scale (ICLR 2024 · similarity 0.87)
- DistillNeRF: Perceiving 3D Scenes from Single-Glance Images by Distilling Neural Fields and Foundation Model Features (NeurIPS 2024 · similarity 0.87)
- From Thousands to Billions: 3D Visual Language Grounding via Render-Supervised Distillation from 2D VLMs (ICML 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.
