Research map / 3D vision / Face and shape
Learning Shape Templates With Structured Implicit Functions
Kyle Genova, Forrester Cole, Daniel Vlasic, Aaron Sarna, William T. Freeman, Thomas Funkhouser
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 structured implicit template instead of voxel grid (3D-R2N2) | 3D-R2N2 · ShapeNet single-view reconstruction (mean, threshold tau) | F-score (%): 39 -> 48 | Improved |
| use structured implicit template instead of voxel grid (3D-R2N2) | 3D-R2N2 · ShapeNet single-view reconstruction (mean, threshold 2 tau) | F-score (%): 55 -> 70 | Improved |
| use structured implicit template instead of point set (PSG) | PSG · ShapeNet single-view reconstruction (mean, threshold tau) | F-score (%): 49 -> 48 | No clear effect |
| use structured implicit template instead of point set (PSG) | PSG · ShapeNet single-view reconstruction (mean, threshold 2 tau) | F-score (%): 70 -> 70 | No clear effect |
| use structured implicit template instead of Pix2Mesh mesh | Pix2Mesh · ShapeNet single-view reconstruction (mean, threshold tau) | F-score (%): 60 -> 48 | Hurt |
| use structured implicit template instead of Pix2Mesh mesh | Pix2Mesh · ShapeNet single-view reconstruction (mean, threshold 2 tau) | F-score (%): 74 -> 70 | Hurt |
| use structured implicit template instead of Pix2Mesh mesh | Pix2Mesh · ShapeNet bench (threshold 2 tau) | F-score (%): 72 -> 82 | Improved |
| use structured implicit template instead of Pix2Mesh mesh | Pix2Mesh · ShapeNet table (threshold tau) | F-score (%): 66 -> 40 | Hurt |
| decode with an LSTM predicting each shape element in succession instead of one fully connected layer (ablation) | feed-forward CNN template network · ShapeNet 3D-to-template fitting | fit quality | Mixed |
| replace the 5-layer CNN encoder with ResNet (ablation) | depth-image template network · ShapeNet 3D-to-template fitting | convergence and generalization | Hurt |
| use ResNet V2 50 with ImageNet pretraining instead of the depth-image CNN for RGB input (ablation) | RGB-to-template network · ShapeNet RGB to template | fit quality | Improved |
Closest papers by meaning
- Deep Implicit Templates for 3D Shape Representation (CVPR 2021 · similarity 0.91)
- Implicit Surface Representations As Layers in Neural Networks (ICCV 2019 · similarity 0.89)
- Learning Manifold Patch-Based Representations of Man-Made Shapes (ICLR 2021 · similarity 0.88)
- ImplicitAtlas: Learning Deformable Shape Templates in Medical Imaging (CVPR 2022 · similarity 0.88)
- Local Deep Implicit Functions for 3D Shape (CVPR 2020 · similarity 0.88)
- Self-supervised Learning of Implicit Shape Representation with Dense Correspondence for Deformable Objects (ICCV 2023 · 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.
