Research map / 3D vision / 3D generation
Learning from Unlabelled Videos Using Contrastive Predictive Neural 3D Mapping
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 view-contrastive rather than RGB regression pre-training | 3D detector trained on KITTI labels · KITTI validation | mAP@IOU 0.50: .55 -> .60 | Helped |
| Fine-tune the pretrained view-contrastive network rather than freeze it (ablation) | 3D detector with CARLA pre-training · KITTI validation | mAP@IOU 0.50: .58 -> .60 | Helped |
| Use view-contrastive rather than RGB regression pre-training | 3D flow estimator · 3D flow: Moving | mean endpoint error: 5.55 -> 3.33 | Helped |
| Add a 3D contrastive loss to the 2D contrastive objective (ablation) | GRNN features · 3D patch retrieval | P@1: .52 -> .80 | Helped |
| Add a 3D contrastive loss to the 2D contrastive objective (ablation) | GRNN features · 2D patch retrieval | P@1: .14 -> .20 | Helped |
| Use distance-weighted negative sampling rather than random or semi-hard sampling | Contrastive predictive training · CARLA view-contrastive training | performance | Helped |
| Combine synthetic rigid-transformation supervision with temporal feature matching | 3D imagination flow module · CARLA dynamic scenes | 3D flow accuracy | Helped |
| Sample small and zero rigid motions in addition to large motions (ablation) | 3D imagination flow module · Synthetic transformation flow training | motion-range accuracy | Helped |
| Use logarithmic rather than linear depth spacing when projecting features (ablation) | 3D-to-2D projection module · CARLA view prediction | projection effectiveness | Helped |
Closest papers by meaning
- CoCoNets: Continuous Contrastive 3D Scene Representations (CVPR 2021 · similarity 0.92)
- Neural Groundplans: Persistent Neural Scene Representations from a Single Image (ICLR 2023 · similarity 0.88)
- OSN: Infinite Representations of Dynamic 3D Scenes from Monocular Videos (ICML 2024 · similarity 0.87)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.87)
- The 3D-PC: a benchmark for visual perspective taking in humans and machines (ICLR 2025 · similarity 0.87)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · 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.