Research map / 3D vision / 3D generation
Towards Scalable Spatial Intelligence via 2D-to-3D Data Lifting
Xingyu Miao, Haoran Duan, Quanhao Qian, Jiuniu Wang, Yang Long, Ling Shao, Deli Zhao, Ran Xu, Gongjie Zhang
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 |
|---|---|---|---|
| pretrain on COCO-3D before fine-tuning on ScanNet | Uni3D+Mask3D · ScanNet instance segmentation | mAP (%): 24.30 -> 28.64 | Improved |
| pretrain on COCO-3D before fine-tuning on ScanNet | Uni3D+Mask3D · ScanNet instance segmentation | mAP@0.25: 67.28 -> 67.33 | No clear effect |
| pretrain on COCO-3D before fine-tuning on ScanNet | SpUNet · ScanNet semantic segmentation | mIoU (%): 31.09 -> 62.48 | Improved |
| pretrain on COCO-3D before fine-tuning on ScanNet | PTv2 · ScanNet semantic segmentation | mIoU (%): 51.04 -> 55.81 | Improved |
| pretrain on COCO-3D before fine-tuning on ScanNet | Uni3D · ScanNet semantic segmentation | mIoU (%): 52.14 -> 55.83 | Improved |
| pretrain on COCO-3D before fine-tuning on ScanNet | TGNN · Referring instance segmentation (ScanNet) | mIoU (%): 26.10 -> 32.47 | Improved |
| pretrain on COCO-3D before fine-tuning on ScanNet | TGNN · Referring instance segmentation (ScanNet) | Acc@0.5: 29.00 -> 37.12 | Improved |
| pretrain on COCO-3D | LL3DA · ScanQA (val) | CIDEr: 76.8 -> 85.2 | Improved |
| pretrain on COCO-3D | LL3DA · ScanRefer dense captioning | C@0.5: 62.98 -> 67.04 | Improved |
| pretrain on COCO-3D | LL3DA · Nr3D dense captioning | C@0.5: 23.94 -> 25.20 | Improved |
| train on COCO-3D only and test zero-shot on ScanNet | TGNN · Referring instance segmentation | mIoU (%): 26.10 -> 10.10 | Hurt |
Closest papers by meaning
- From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D (NeurIPS 2025 · similarity 0.88)
- HiSpatial: Taming Hierarchical 3D Spatial Understanding in Vision-Language Models (CVPR 2026 · similarity 0.88)
- Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering (NeurIPS 2026 · similarity 0.88)
- VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction (CVPR 2026 · similarity 0.88)
- Abstract 3D Perception for Spatial Intelligence in Vision-Language Models (CVPR 2026 · similarity 0.88)
- Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence (NeurIPS 2025 · similarity 0.88)
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.
