Research map / 3D vision / Face and shape
Cuboids Revisited: Learning Robust 3D Shape Fitting to Single RGB Images
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the proposed robust cuboid fitting method instead of SQ-Parsing + BTS | RGB scene parsing with BTS features · NYU Depth v2 test | AUC@20cm (OA-L2): 11.6% -> 33.1% | Helped |
| Use the proposed robust cuboid fitting method instead of SQ-Parsing + BTS | RGB scene parsing with BTS features · NYU Depth v2 test | mean OA-L2 (cm): 65.9 -> 34.5 | Helped |
| Use the proposed robust cuboid fitting method instead of SQ-Parsing + BTS | RGB scene parsing with BTS features · NYU Depth v2 test | mean L2 (cm): 33.9 -> 30.1 | Helped |
| Use the proposed robust cuboid fitting method instead of Sequential RANSAC | depth-input scene parsing · NYU Depth v2 test | AUC@20cm (OA-L2): 42.8% -> 62.7% | Helped |
| Use the proposed robust cuboid fitting method instead of SQ-Parsing | depth-input scene parsing · NYU Depth v2 test | mean OA-L2 (cm): 34.7 -> 20.8 | Helped |
| Use the proposed robust cuboid fitting method instead of SQ-Parsing | depth-input scene parsing · NYU Depth v2 test | mean L2 (cm): 20.9 -> 17.9 | Helped |
| Increase sampling-weight sets from Q=1 to Q=4 (ablation) | the proposed method · NYU Depth v2 test, depth input | AUC@20cm (OA-L2): 40.2% -> 62.7% | Helped |
| Increase sampling-weight sets from Q=2 to Q=4 (ablation) | the proposed method · NYU Depth v2 test, depth input | AUC@20cm (OA-L2): 57.5% -> 62.7% | Helped |
| Enable occlusion-aware counting at inference rather than disabling it (ablation) | the proposed method · NYU Depth v2 test, depth input | AUC@20cm (OA-L2): 5.3% -> 62.7% | Helped |
| Enable occlusion-aware counting for training sampling and selection (ablation) | the proposed method · NYU Depth v2 test, depth input | AUC@20cm (OA-L2): 58.8% -> 62.7% | Helped |
| Enable occlusion-aware counting in the training loss (ablation) | the proposed method · NYU Depth v2 test, depth input | AUC@20cm (OA-L2): 61.8% -> 62.7% | Helped |
| Enable occlusion-aware training sampling and loss together (ablation) | the proposed method · NYU Depth v2 test, depth input | AUC@20cm (OA-L2): 61.2% -> 62.7% | Helped |
Closest papers by meaning
- Im2Struct: Recovering 3D Shape Structure From a Single RGB Image (CVPR 2018 · similarity 0.89)
- Learning Unsupervised Hierarchical Part Decomposition of 3D Objects From a Single RGB Image (CVPR 2020 · similarity 0.88)
- Cubify Anything: Scaling Indoor 3D Object Detection (CVPR 2025 · similarity 0.88)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.87)
- EA3D: Online Open-World 3D Object Extraction from Streaming Videos (NeurIPS 2025 · similarity 0.86)
- Common3D: Self-Supervised Learning of 3D Morphable Models for Common Objects in Neural Feature Space (CVPR 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.