Research map / 3D vision / 3D generation
Associative3D: Volumetric Reconstruction from Sparse Views
Shengyi Qian, Linyi Jin, David F. Fouhey
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 stitching stage with joint reasoning over objects and cameras (ablation) | Feedforward · SUNCG test | AP (all errors): 21.2 -> 23.3 | Improved |
| add stitching stage (ablation) | Feedforward · SUNCG test | AP (translation only): 31.7 -> 38.4 | Improved |
| add stitching stage (ablation) | Feedforward · SUNCG test (top 25%) | AP (all errors): 41.6 -> 48.3 | Improved |
| merge duplicate objects with NMS (ablation) | Feedforward · SUNCG test | AP (all errors): 21.2 -> 21.1 | No clear effect |
| merge objects by raw top-1 affinity instead of stitching (ablation) | Feedforward · SUNCG test | AP (all errors): 21.2 -> 15.0 | Hurt |
| add stitching stage to affinity-based correspondence (ablation) | affinity top-1 · SUNCG test object correspondence | AP: 49.4 -> 60.0 | Improved |
| use top-1 affinity with stitching confidence instead of raw affinity (ablation) | affinity · SUNCG test object correspondence | AP: 38.8 -> 49.4 | Improved |
| pick camera pose by stitching rather than top-1 camera branch prediction (ablation) | relative camera pose · SUNCG test | translation error <= 1m (%): 41.26 -> 54.89 | Improved |
| pick camera pose by stitching rather than top-1 (ablation) | relative camera pose · SUNCG test | rotation error <= 30 deg (%): 77.56 -> 78.31 | No clear effect |
| use two views instead of a single view | 3D-RelNet vs Associative3D · SUNCG test | AP (all errors): 13.7 -> 23.3 | Improved |
| average rotation of matched objects instead of random pick (ablation) | stitching stage · SUNCG test (top 50%) | AP (all errors): 38.8 -> 31.4 | Hurt |
| average shape vector of matched objects instead of random pick (ablation) | stitching stage · SUNCG test (top 50%) | AP (all errors): 38.8 -> 38.8 | No clear effect |
Closest papers by meaning
- CoReNet: Coherent 3D Scene Reconstruction from a Single RGB Image (ECCV 2020 · similarity 0.89)
- Semantic Multi-View Stereo: Jointly Estimating Objects and Voxels (CVPR 2017 · similarity 0.88)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.87)
- Know Your Neighbors: Improving Single-View Reconstruction via Spatial Vision-Language Reasoning (CVPR 2024 · similarity 0.87)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.87)
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.87)
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.
