Research map / 3D vision / 3D generation
InstaScene: Towards Complete 3D Instance Decomposition and Reconstruction from Cluttered Scenes
Zesong Yang, Bangbang Yang, Wenqi Dong, Chenxuan Cao, Liyuan Cui, Yuewen Ma, Zhaopeng Cui, Hujun Bao
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 filtered intra-view 2D masks to contrastive learning on 3D masks (ablation) | InstaScene spatial contrastive learning · LERF-Mask (average) | mIoU (%): 79.0 -> 83.6 | Improved |
| add cross-view 2D masks on top of filtered 2D masks (ablation) | InstaScene spatial contrastive learning · LERF-Mask (average) | mIoU (%): 83.6 -> 85.6 | Improved |
| use 3D mask guidance with filtered and cross-view masks instead of raw noisy 2D masks (ablation) | InstaScene · LERF-Mask (average) | mIoU (%): 80.5 -> 85.6 | Improved |
| use coarse 3D masks alone instead of noisy 2D masks (ablation) | InstaScene · LERF-Mask (average) | mIoU (%): 80.5 -> 79.0 | Hurt |
| InstaScene versus LangSplat and GSGrouping for instance segmentation | InstaScene · LERF-Mask (average) | mIoU (%): 60.6 -> 85.6 | Improved |
| in-situ generation versus SpaRP multi-view completion | InstaScene · Replica-CAD instance completion | Volume IoU: 0.590 -> 0.716 | Improved |
| in-situ generation versus original 2DGS reconstruction on unknown views | InstaScene · Replica-CAD instance completion | PSNR (unknown views): 27.44 -> 29.02 | Improved |
| in-situ generation versus 2DGS on Chamfer Distance | InstaScene · Replica-CAD instance completion | Chamfer Distance: 0.028 -> 0.016 | Improved |
| alternate conditioning views instead of single-view conditioning (ablation) | InstaScene in-situ generation · ZipNeRF / Replica-CAD (qualitative) | visual quality | Improved |
| add geometry-aware warped feature conditioning to alternated views (ablation) | InstaScene in-situ generation · ZipNeRF / Replica-CAD (qualitative) | visual quality | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- MVDFusion — MVD-Fusion: Single-view 3D via Depth-consistent Multi-view Generation (CVPR 2024) · better on 5 of 5 shared results
- SpaRP — SpaRP: Fast 3D Object Reconstruction and Pose Estimation from Sparse Views (ECCV 2024) · better on 5 of 5 shared results
- LangSplat — LangSplat: 3D Language Gaussian Splatting (CVPR 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Inst3D-LMM: Instance-Aware 3D Scene Understanding with Multi-modal Instruction Tuning (CVPR 2025 · similarity 0.87)
- InstaInpaint: Instant 3D-Scene Inpainting with Masked Large Reconstruction Model (NeurIPS 2025 · similarity 0.87)
- Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering (NeurIPS 2026 · similarity 0.86)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.86)
- InfiniBench: Infinite Benchmarking for Visual Spatial Reasoning with Customizable Scene Complexity (CVPR 2026 · similarity 0.86)
- UniSDF: Unifying Neural Representations for High-Fidelity 3D Reconstruction of Complex Scenes with Reflections (NeurIPS 2024 · 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.
