Research map / 3D vision / Face and shape
Escaping Plato's Cave: 3D Shape From Adversarial Rendering
Philipp Henzler, Niloy J. Mitra, Tobias Ritschel
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 Platonic 2D adversarial rendering instead of PrGAN | PrGAN · ShapeNet airplane (VH) | FID (lower better): 207 -> 55 | Improved |
| use Platonic 2D adversarial rendering instead of PrGAN | PrGAN · ShapeNet airplane (AO) | FID (lower better): 190 -> 90 | Improved |
| use Platonic 2D adversarial rendering instead of PrGAN | PrGAN · ShapeNet airplane (EA) | FID (lower better): 196 -> 181 | Improved |
| use Platonic 2D adversarial rendering instead of PrGAN | PrGAN · ShapeNet airplane (EA) | IoU: 0.08 -> 0.13 | Hurt |
| add Platonic adversarial rendering loss to 3DGAN | 3DGAN · ShapeNet airplane (VH) | FID (lower better): 111 -> 98 | Improved |
| add Platonic adversarial rendering loss to 3DGAN | 3DGAN · ShapeNet airplane (AO) | FID (lower better): 58 -> 64 | Hurt |
| add Platonic adversarial rendering loss to 3DGAN | 3DGAN · ShapeNet airplane (EA) | FID (lower better): 110 -> 73 | Improved |
| add Platonic adversarial rendering loss to 3DGAN | 3DGAN · ShapeNet airplane (all IF) | 3D metrics (IoU, RMSE): 0.46 -> 0.44 IoU (VH) | Hurt |
| add 70k 2D images to 3D GAN training with 5 3D shapes (ablation) | 3DGAN · ShapeNet (5 3D shapes) | mean DSSIM (lower better): .135 -> .134 | Improved |
| add 70k 2D images to 3D GAN training with 250 3D shapes (ablation) | 3DGAN · ShapeNet (250 3D shapes) | mean DSSIM (lower better): .106 -> .102 | Improved |
| add 70k 2D images to 3D GAN training with 1.5k 3D shapes (ablation) | 3DGAN · ShapeNet (1.5k 3D shapes) | mean DSSIM (lower better): .101 -> .099 | Improved |
| add 70k 2D images to 3D GAN training with 50 3D shapes (ablation) | 3DGAN · ShapeNet (50 3D shapes) | mean DSSIM (lower better): .108 -> .108 | No clear effect |
Closest papers by meaning
- Do 2D GANs Know 3D Shape? Unsupervised 3D Shape Reconstruction from 2D Image GANs (ICLR 2021 · similarity 0.89)
- Inferring 3D Shapes from Image Collections Using Adversarial Networks (IJCV 2020 · similarity 0.89)
- From Images to 3D Shape Attributes (TPAMI 2017 · similarity 0.88)
- Unsupervised Learning of 3D Structure from Images (NeurIPS 2016 · similarity 0.88)
- Geometry-Aware Network for Non-Rigid Shape Prediction From a Single View (CVPR 2018 · similarity 0.87)
- HoloGAN: Unsupervised Learning of 3D Representations From Natural Images (ICCV 2019 · 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.
