Research map / 3D vision / Face and shape
Monocular 3D Object Reconstruction with GAN Inversion
Junzhe Zhang, Daxuan Ren, Zhongang Cai, Chai Kiat Yeo, Bo Dai, Chen Change Loy
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 Chamfer mask loss instead of IoU mask loss (ablation) | MeshInversion · CUB | IoU: 0.604 -> 0.752 | Improved |
| use Chamfer mask loss instead of L1 mask loss (ablation) | MeshInversion · CUB | IoU: 0.589 -> 0.752 | Improved |
| use Chamfer texture loss instead of L1 texture loss (ablation) | MeshInversion · CUB | FID_12: 74.7 -> 56.8 | Improved |
| use Chamfer texture loss instead of L1 + perceptual loss (ablation) | MeshInversion · CUB | FID_12: 64.9 -> 56.8 | Improved |
| use Chamfer texture loss instead of contextual loss (ablation) | MeshInversion · CUB | FID_12: 75.5 -> 56.8 | Improved |
| use Chamfer texture loss instead of contextual loss (ablation) | MeshInversion · CUB | IoU: 0.718 -> 0.752 | Improved |
| fine-tune camera pose jointly with latent code (ablation) | MeshInversion · CUB | IoU: 0.695 -> 0.752 | Improved |
| fine-tune camera pose jointly with latent code (ablation) | MeshInversion · CUB | FID_12: 60.0 -> 56.8 | Improved |
| add image-space discriminator in GAN pre-training (ablation) | ConvMesh · CUB (pre-training) | Full FID: 33.6 -> 28.3 | Improved |
| lower epsilon_s (more tolerance to local misalignment) (ablation) | Chamfer texture loss · CUB | FID_12: 58.0 -> 56.8 | Improved |
| add test-time optimization to CMR | CMR · CUB | FID_12: 180.1 -> 158.4 | Improved |
| add test-time optimization to UMR | UMR · CUB | FID_12: 86.9 -> 91.3 | Hurt |
Closest papers by meaning
- 3D GAN Inversion With Facial Symmetry Prior (CVPR 2023 · similarity 0.90)
- High-Fidelity 3D GAN Inversion by Pseudo-Multi-View Optimization (CVPR 2023 · similarity 0.90)
- One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimization (NeurIPS 2023 · similarity 0.90)
- WarpGAN: Warping-Guided 3D GAN Inversion with Style-Based Novel View Inpainting (NeurIPS 2025 · similarity 0.88)
- Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images (NeurIPS 2024 · similarity 0.87)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · 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.
