Research map / Point clouds / Shape and surface completion
Multi-view 3D Models from Single Images with a Convolutional Network
Maxim Tatarchenko, Alexey Dosovitskiy, Thomas Brox
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 |
|---|---|---|---|
| use the encoder-decoder network instead of nearest-neighbor HOG baseline | unseen view prediction · ShapeNet normal test cars | color error: 0.028 -> 0.013 | Improved |
| use the encoder-decoder network instead of nearest-neighbor RGB baseline | unseen view prediction · ShapeNet normal test cars | color error: 0.018 -> 0.013 | Improved |
| use the network instead of nearest-neighbor HOG+RGB baseline | unseen view prediction · ShapeNet difficult test cars | color error: 0.036 -> 0.028 | Improved |
| use the network instead of nearest-neighbor RGB baseline | unseen view prediction · ShapeNet difficult test cars | depth error: 0.0265 -> 0.0207 | Improved |
| use the network instead of nearest-neighbor HOG baseline | unseen view prediction · ShapeNet normal test cars | depth error: 0.0058 -> 0.0057 | No clear effect |
| use the network instead of Dosovitskiy et al. | unseen view prediction · ShapeNet chairs test | average error: 0.0308 -> 0.0208 | Improved |
| render training data realistically (random lights, alpha compositing, smoothing) instead of basic rendering (ablation) | multi-view 3D network · real car images | qualitative shape estimation | Improved |
| add adversarial loss (alpha=0.01) to squared Euclidean loss (ablation) | multi-view 3D network · ShapeNet cars | image sharpness and shape | Mixed |
| add data augmentation (ablation) | multi-view 3D network · ShapeNet cars | generalization | No clear effect |
| use deeper and wider networks (ablation) | multi-view 3D network · ShapeNet cars | performance | No clear effect |
Closest papers by meaning
- MVPNet: Multi-View Point Regression Networks for 3D Object Reconstruction from A Single Image (AAAI 2019 · similarity 0.88)
- Learning to Generate Chairs, Tables and Cars with Convolutional Networks (TPAMI 2016 · similarity 0.88)
- 3D-RCNN: Instance-Level 3D Object Reconstruction via Render-and-Compare (CVPR 2018 · similarity 0.88)
- DeepMVS: Learning Multi-View Stereopsis (CVPR 2018 · similarity 0.87)
- Fully-Convolutional Point Networks for Large-Scale Point Clouds (ECCV 2018 · similarity 0.87)
- Capturing the Geometry of Object Categories from Video Supervision (TPAMI 2018 · 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.
