Research map / 3D vision / Face and shape
AvatarMe++: Facial Shape and BRDF Inference With Photorealistic Rendering-Aware GANs
Alexandros Lattas, Stylianos Moschoglou, Stylianos Ploumpis, Barış Geçer, Abhijeet Ghosh, Stefanos P. Zafeiriou
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 rendering loss (ablation) | single-network AvatarMe++ · test set T reflectance | MSE: 0.0080 -> 0.0075 | Improved |
| add whole low-res texture input (ablation) | single-network AvatarMe++ · test set T reflectance | MSE: 0.0075 -> 0.0066 | Improved |
| add random rendering environment (ablation) | AvatarMe++ · test set T reflectance | MSE: 0.0066 -> 0.0055 | Improved |
| add multiple random environments (ablation) | AvatarMe++ · test set T reflectance | MSE: 0.0055 -> 0.0048 | Improved |
| add self-occlusion autoencoder (ablation) | AvatarMe++ · test set T reflectance | MSE: 0.0048 -> 0.0043 | Improved |
| replace four networks with a single network (ablation) | AvatarMe · test set T reflectance | MSE: 0.0079 -> 0.0080 | No clear effect |
| add 50%/25%/25% scene variation in rendering loss (ablation) | AvatarMe++ · test set T | Recon. MSE: 0.0055 -> 0.0043 | Improved |
| increase scene variation to 75%/50%/50% (ablation) | AvatarMe++ · test set T | Recon. MSE: 0.0043 -> 0.0051 | Hurt |
| use 2 rendering scenes instead of 1 (ablation) | AvatarMe++ · test set T | Recon. MSE: 0.0054 -> 0.0045 | Improved |
| use 3 rendering scenes instead of 2 (ablation) | AvatarMe++ · test set T | Recon. MSE: 0.0045 -> 0.0043 | Improved |
| set rendering loss weight lambda_R to 0.3 (ablation) | AvatarMe++ · test set T | reflectance MSE | Improved |
| upgrade AvatarMe to AvatarMe++ versus prior methods | AvatarMe++ · in-the-wild reconstructions | PSNR (Albedo): 24.05 -> 26.18 | Improved |
Later papers that compared with this method
1 later accepted paper reports AvatarMe++ as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Relightify: Relightable 3D Faces from a Single Image via Diffusion Models (ICCV 2023 · better on 2 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- AvatarMe: Realistically Renderable 3D Facial Reconstruction "In-the-Wild" (CVPR 2020 · similarity 0.96)
- FitMe: Deep Photorealistic 3D Morphable Model Avatars (CVPR 2023 · similarity 0.90)
- Normalized Avatar Synthesis Using StyleGAN and Perceptual Refinement (CVPR 2021 · similarity 0.90)
- GANFIT: Generative Adversarial Network Fitting for High Fidelity 3D Face Reconstruction (CVPR 2019 · similarity 0.89)
- High-Fidelity 3D Head Avatars Reconstruction through Spatially-Varying Expression Conditioned Neural Radiance Field (AAAI 2024 · similarity 0.89)
- DiffHuman: Probabilistic Photorealistic 3D Reconstruction of Humans (CVPR 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.
