Research map / 3D vision / 3D generation
TriO: Tri-Modal Unsupervised Occupancy World Model for Anything Perception
Quinlan Sykora, Sourav Biswas, Christopher Diehl, Andrew Cunningham, Thomas Gilles, Raquel Urtasun
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 unsupervised pre-training (ablation) | TriO · LRR (200m+) | mAP: 41.0 -> 42.6 | Improved |
| add RADAR as input modality (ablation) | TriO · LRR (200m+) | mAP: 34.5 -> 42.6 | Improved |
| add camera input (ablation) | TriO · LRR (200m+) | mAP: 42.2 -> 42.6 | No clear effect |
| add camera input (ablation) | TriO · LRR (all ranges) | EPE: 2.68 -> 2.86 | Hurt |
| add RADAR Doppler flow loss (ablation) | TriO · LRR (all ranges) | EPE: 4.35 -> 2.86 | Improved |
| add RADAR Doppler flow loss (ablation) | TriO · LRR (200m+) | EPE: 6.02 -> 4.01 | Improved |
| use soft probabilistic labels instead of hard thresholded labels (CSL Soft vs Hard) | TriO · STU semantic occupancy | AP: 76.1 -> 81.6 | Improved |
| add camera-based image supervision to LiDAR-only model under CSL labels (TriO vs DiO) | TriO · STU semantic occupancy | mIoU: 49.8 -> 54.9 | Improved |
| fuse image and LiDAR pseudo-labels (CSL vs image-only ISL) | pseudo-labels · STU pseudo-label segmentation | mIoU: 38.6 -> 44.6 | Improved |
| add oracle class prompts to OWLv2 + SAM | OWLv2 + SAM · STU pseudo-label segmentation | mIoU: 20.2 -> 28.6 | Improved |
| pre-train with RADAR in TriO vs DIO on flow in AV2 | TriO · AV2 fine-tuned occupancy | EPE: 1.43 -> 1.97 | Hurt |
Closest papers by meaning
- UnO: Unsupervised Occupancy Fields for Perception and Forecasting (CVPR 2024 · similarity 0.86)
- OpenAD: Open-World Autonomous Driving Benchmark for 3D Object Detection (NeurIPS 2025 · similarity 0.86)
- Learning Occupancy for Monocular 3D Object Detection (CVPR 2024 · similarity 0.86)
- SDGOCC: Semantic and Depth-Guided Bird's-Eye View Transformation for 3D Multimodal Occupancy Prediction (CVPR 2025 · similarity 0.86)
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 2025 · similarity 0.86)
- CSV-Occ: Fusing Multi-frame Alignment for Occupancy Prediction with Temporal Cross State Space Model and Central Voting Mechanism (ICML 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.
