Research map / 3D vision / Face and shape
Deep Meta Functionals for Shape Representation
Gidi Littwin, Lior Wolf
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 |
|---|---|---|---|
| sample 3D training points near the object boundary instead of uniformly (ablation) | deep meta functionals · ShapeNet-core | mean IOU (%): 63.5 -> 65.8 | Improved |
| use larger number of ResNet blocks B=5 instead of B=4 in f (ablation) | deep meta functionals · ShapeNet-core | IOU (%): 67.0 -> 67.3 | Improved |
| change N from 64 to 128 base kernels while setting K from 2 to 0 (ablation) | deep meta functionals · ShapeNet-core | IOU (%): 67.3 -> 67.3 | No clear effect |
| vary activation (ELU, ReLU, tanh) and depth/width of g (ablation) | deep meta functionals · ShapeNet-core | IOU (%): 64.5 -> 66.1 | No clear effect |
| average embeddings of multiple views at test time | deep meta functionals · ShapeNet-core 32^3 | mean IOU (%) | Improved |
| use deep meta functionals instead of PCDI (best prior at 32^3) | single-image reconstruction · ShapeNet-core 32^3 | mean IOU (%): 64.8 -> 69.1 | Improved |
| use deep meta functionals instead of VP3D at 256^3 | single-image reconstruction · ShapeNet-core 256^3 | mean IOU (%): 65.5 -> 66.2 | Improved |
| use deep meta functionals instead of AtlasNet | single-image reconstruction · ShapeNet-core 256^3 | average CD x10^3: 9.52 -> 4.35 | Improved |
| train from three silhouettes with differentiable rendering | deep meta functionals · 3-silhouette learning | IOU: 60.0 -> 64.4 | Improved |
Closest papers by meaning
- MetaSDF: Meta-Learning Signed Distance Functions (NeurIPS 2020 · similarity 0.89)
- Image-Based 3D Object Reconstruction: State-of-the-Art and Trends in the Deep Learning Era (TPAMI 2019 · similarity 0.88)
- From Images to 3D Shape Attributes (TPAMI 2017 · similarity 0.88)
- OperatorNet: Recovering 3D Shapes From Difference Operators (ICCV 2019 · similarity 0.88)
- PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representations (ECCV 2020 · similarity 0.88)
- Local Deep Implicit Functions for 3D Shape (CVPR 2020 · similarity 0.88)
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.
