Research map / 3D vision / 3D generation
nuCraft: Crafting High Resolution 3D Semantic Occupancy for Unified 3D Scene Understanding
Benjin Zhu, zhe wang, Hongsheng Li
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 |
|---|---|---|---|
| encode occupancy GT with a VQ-VAE and predict by simulating latents (VQ-Occ) instead of direct prediction | VQ-Occ · OpenOccupancy val (camera+LiDAR) | mIoU: 20.1 -> 25.5 (vs M-CONet) | Improved |
| use VQ-Occ instead of C-CONet with camera input on OpenOccupancy | VQ-Occ · OpenOccupancy val (camera) | mIoU: 12.8 -> 13.6 | Improved |
| use VQ-Occ at 0.1m directly instead of upsampling a coarse C-CONet prediction | VQ-Occ · nuCraft@0.1 (camera, batch size 1) | mIoU: 3.8 -> 9.6 | Improved |
| increase occupancy GT resolution from 0.2m to 0.1m for detection on occupancy centroids (ablation) | nuCraft · Voxel-DETR 3D detection on occupancy input | mAP: 62.7 -> 77.9 | Improved |
| use nuCraft GT instead of OpenOccupancy for drivable area | nuCraft · Drivable area segmentation (HD map) | mIoU: 64.1 -> 69.8 (0.2m) | Improved |
| add mesh reconstruction in nuCraft data generation (ablation) | nuCraft pipeline · nuCraft@0.2 data generation | mIoU: 63.2 -> 69.8 | Improved |
| add pose estimation (Kiss-ICP) in LiDAR aggregation (ablation) | nuCraft pipeline · nuCraft@0.2 data generation | mIoU: 66.1 -> 69.8 | Improved |
| add continuous scenes grouping (ablation) | nuCraft pipeline · nuCraft@0.2 data generation | mIoU: 66.5 -> 69.8 | Improved |
| add static/moving parts separation (ablation) | nuCraft pipeline · nuCraft@0.2 data generation | mIoU: 67.2 -> 69.8 | Improved |
| add mesh cleaning (ablation) | nuCraft pipeline · nuCraft@0.2 data generation | mIoU: 69.7 -> 69.8 | No clear effect |
| train and test across OpenOccupancy and nuCraft instead of same dataset | VQ-Occ (camera) · Cross-dataset (camera, 0.2m) | mIoU: -3.9 (OO to nC, 9.7); -4.8 (nC to OO, 9.3) | Hurt |
Closest papers by meaning
- OctOcc: High-Resolution 3D Occupancy Prediction with Octree (AAAI 2024 · similarity 0.94)
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 2025 · similarity 0.93)
- CSV-Occ: Fusing Multi-frame Alignment for Occupancy Prediction with Temporal Cross State Space Model and Central Voting Mechanism (ICML 2025 · similarity 0.92)
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 2025 · similarity 0.91)
- LowRankOcc: Tensor Decomposition and Low-Rank Recovery for Vision-based 3D Semantic Occupancy Prediction (CVPR 2024 · similarity 0.90)
- QuadricFormer: Scene as Superquadrics for 3D Semantic Occupancy Prediction (NeurIPS 2025 · similarity 0.90)
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.
