Research map / 3D vision / 3D generation
Tracking Emerges by Looking Around Static Scenes, with Neural 3D Mapping
Adam W. Harley, Shrinidhi Kowshika Lakshmikanth, Paul Schydlo, Katerina Fragkiadaki
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 search regions centered on last known position (ablation) | 3D neural mapping tracker · CARLA | mean 3D IOU: 0.40 -> 0.61 | Improved |
| use full resolution instead of half resolution without search regions (ablation) | 3D neural mapping tracker · CARLA | mean 3D IOU: 0.25 -> 0.40 | Improved |
| add static point selection with reliability network g (ablation) | 3D neural mapping tracker · KITTI | mean 3D IOU: 0.39 -> 0.46 | Improved |
| learn features with multiview contrastive loss instead of random features | 3D neural mapping · CARLA IOU@8 | mean 3D IOU: 0.03 -> 0.61 | Improved |
| use 3D neural map instead of 2.5D dense object nets | tracker · CARLA IOU@8 | mean 3D IOU: 0.33 -> 0.61 | Improved |
| use 3D neural map instead of 2.5D dense object nets | tracker · KITTI IOU@8 | mean 3D IOU: 0.31 -> 0.46 | Improved |
| use contrastive static-point loss instead of colorization | 2.5D tracker · CARLA IOU@8 | mean 3D IOU: 0.19 -> 0.33 | Improved |
| use our unsupervised model instead of unsupervised 3D flow | tracker · KITTI IOU@8 | mean 3D IOU: 0.22 -> 0.46 | Improved |
| add cosine windowing | 3D siamese supervised tracker · CARLA IOU@8 | mean 3D IOU: 0.61 -> 0.65 | Improved |
| add cosine windowing | 3D siamese supervised tracker · KITTI IOU@8 | mean 3D IOU: 0.58 -> 0.55 | Hurt |
| use supervised siamese + cosine instead of unsupervised ours | tracker · CARLA IOU@8 | mean 3D IOU: 0.65 vs 0.61 | Mixed |
| track static objects instead of moving objects | Ours · CARLA IOU@8 | mean 3D IOU: 0.64 vs 0.56 | Mixed |
Closest papers by meaning
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · similarity 0.89)
- Seeing Behind Objects for 3D Multi-Object Tracking in RGB-D Sequences (CVPR 2021 · similarity 0.88)
- Learning from Unlabelled Videos Using Contrastive Predictive Neural 3D Mapping (ICLR 2020 · similarity 0.88)
- Multi-View 3D Point Tracking (ICCV 2025 · similarity 0.88)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.87)
- TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels (NeurIPS 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.
