Research map / 3D vision / Face and shape
Progressive Learning of 3D Reconstruction Network From 2D GAN Data
Ayşegül Dündar, Jun Gao, Andrew Tao, Bryan Catanzaro
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 second stage same-view reconstruction loss (ablation) | Stage I model · Car (novel view) | FID: 11.98 -> 9.05 | Improved |
| add third stage adversarial training (ablation) | Stage II model · Car (novel view) | FID: 9.05 -> 6.92 | Improved |
| add third stage adversarial training (ablation) | Stage II model · Bird (same view) | FID: 64.67 -> 61.21 | Improved |
| add third stage adversarial training (ablation) | Stage II model · Horse (novel view) | FID: 67.77 -> 67.30 | Improved |
| add multi-stage training instead of training from scratch with the final loss (ablation) | Stage III model · Car (novel view) | FID: 66.71 -> 6.92 | Improved |
| train progressively instead of from scratch with the second-stage objective (ablation) | Stage II model · Car (novel view) | FID: 10.00 -> 9.05 | Improved |
| use novel-view feedback instead of same-view single-view training (ablation) | Stage III model · Car (novel view) | FID: 42.24 -> 6.92 | Improved |
| same-view training gives better same-view IoU (ablation) | single-view training · Car (same view) | IoU: 0.94 -> 0.96 | Improved |
| add multi-view conditioning to discriminator (ablation) | Stage III model · Car (novel view) | FID: 7.02 -> 6.92 | Improved |
| add Gaussian latent sampling (ablation) | Stage III model · Car (novel view) | FID: 7.05 -> 6.92 | Improved |
| use our Stage I versus GanVerse trained on same data | Stage I model · Car (same view) | FID: 28.04 -> 10.75 | Improved |
| Stage I versus GanVerse | Stage I model · Horse (same view) | FID: 62.38 -> 83.66 | Hurt |
Closest papers by meaning
- Fine Detailed Texture Learning for 3D Meshes With Generative Models (TPAMI 2023 · similarity 0.91)
- High-Fidelity 3D GAN Inversion by Pseudo-Multi-View Optimization (CVPR 2023 · similarity 0.91)
- 2D GANs Meet Unsupervised Single-View 3D Reconstruction (ECCV 2022 · similarity 0.90)
- Fast-GANFIT: Generative Adversarial Network for High Fidelity 3D Face Reconstruction (TPAMI 2021 · similarity 0.90)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.89)
- Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images (NeurIPS 2024 · 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.
