Research map / 3D vision / Face and shape
Pix2Vox++: Multi-scale Context-aware 3D Object Reconstruction from Single and Multiple Images
Haozhe Xie, Hongxun Yao, Shengping Zhang, Shangchen Zhou, Wenxiu Sun
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 the refiner (ablation) | Pix2Vox++/A · ShapeNet single-view (32^3) | mean IoU: 0.658 -> 0.670 | Improved |
| use multi-scale context-aware fusion instead of average pooling fusion (ablation) | Pix2Vox++/A · ShapeNet-Cars multi-view (128^3), 8 views | mean IoU: 0.829 -> 0.843 | Improved |
| use multi-scale context-aware fusion instead of context-aware fusion (ablation) | Pix2Vox++/A · ShapeNet-Cars multi-view (128^3), 8 views | mean IoU: 0.801 -> 0.843 | Improved |
| use multi-scale context-aware fusion instead of average pooling fusion (ablation) | Pix2Vox++/F · ShapeNet-Cars multi-view (128^3), 8 views | mean IoU: 0.806 -> 0.819 | Improved |
| use multi-scale context-aware fusion instead of context-aware fusion (ablation) | Pix2Vox++/F · ShapeNet-Cars multi-view (128^3), 8 views | mean IoU: 0.768 -> 0.819 | Improved |
| add more input views (1 to 20) to the fusion (ablation) | Pix2Vox++/A · ShapeNet multi-view (32^3) | mean IoU: 0.670 -> 0.719 | Improved |
| swap ResNet50 backbone for VGG16 (pretrained) (ablation) | Pix2Vox++/A · ShapeNet single-view (32^3) | mean IoU: 0.670 -> 0.661 | Hurt |
| swap ResNet50 backbone for DenseNet169 (pretrained) (ablation) | Pix2Vox++/A · ShapeNet single-view (32^3) | mean IoU: 0.670 -> 0.669 | No clear effect |
| use pretrained backbone weights (ablation) | Pix2Vox++/A · ShapeNet single-view (32^3) | mean IoU: 0.669 -> 0.670 | Improved |
| train on Things3D-Chairs instead of ShapeNet-Chairs (ablation) | Pix2Vox++/A · Pix3D chairs (32^3) | mean IoU: 0.204 -> 0.269 | Improved |
| train on Things3D-Chairs-RfC instead of ShapeNet-Chairs-RfC (ablation) | Pix2Vox++/A · Pix3D chairs (32^3) | mean IoU: 0.292 -> 0.324 | Improved |
| replace multi-scale fusion with 3D conv LSTM (A-R2N2) (ablation) | Pix2Vox++/A · ShapeNet multi-view (32^3) | mean IoU | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Pixel2Mesh — Pixel2Mesh: Generating 3D Mesh Models from Single RGB Images (ECCV 2018) · better on 1 of 1 shared result
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
- Pix2Vox: Context-Aware 3D Reconstruction From Single and Multi-View Images (ICCV 2019 · similarity 0.98)
- Pix2Shape: Towards Unsupervised Learning of 3D Scenes from Images Using a View-Based Representation (IJCV 2020 · similarity 0.90)
- Pix3D: Dataset and Methods for Single-Image 3D Shape Modeling (CVPR 2018 · similarity 0.90)
- 3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction (ECCV 2016 · similarity 0.88)
- CoReNet: Coherent 3D Scene Reconstruction from a Single RGB Image (ECCV 2020 · similarity 0.88)
- Object-X: Learning to Reconstruct Multi-Modal 3D Object Representations (NeurIPS 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.
