Research map / 3D vision / Face and shape
Deep Appearance Models: A Deep Boltzmann Machine Approach for Face Modeling
Chi Nhan Duong, Khoa Luu, Kha Gia Quach, Tien Dai Bui
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 DAMs reconstruction instead of bicubic/AAMs | face super-resolution · LFPW (alpha=16) | RMSE: 19.74 -> 19.24 | Improved |
| use DAMs instead of AAMs | face super-resolution · Helen (alpha=16) | RMSE: 22.3 -> 21.24 | Improved |
| use DAMs instead of AAMs | face super-resolution · LFPW (alpha=8) | PSNR (dB): 24.82 -> 25.06 | Improved |
| use DAMs instead of AAMs at small magnification | face super-resolution · Helen (alpha=4) | PSNR (dB): 26.56 -> 26.54 | No clear effect |
| use DAMs frontalization instead of LFW3D | Pittpatt on LFW · LFW verification | accuracy (%): 86.63% -> 87.72% | Improved |
| use DAMs instead of PCA-based AAMs | occlusion removal reconstruction · AR occluded subset | RMSE vs neutral faces: 47.36 -> 45.08 | Improved |
| use DAMs instead of RoBM | occlusion recognition · AR sunglasses | rank-1 (%): 71.89% -> 76.87% | Improved |
| use DAMs instead of AAMs under impulsive noise | age estimation on reconstructed faces · FG-NET (noise range 25) | MAE (years): 6.15 -> 5.81 | Improved |
| use DAMs instead of AAMs for low-resolution faces | age estimation · FG-NET (alpha=8) | MAE (years): 6.69 -> 6.21 | Improved |
| use DAMs model parameters as features instead of AAMs | age estimation · FG-NET | MAE (years): 4.81 -> 4.67 | Improved |
| use DAMs model parameters instead of reconstructed image | age estimation · FG-NET | MAE (years): 5.67 -> 4.67 | Improved |
| use DAMs instead of AAMs for shape fitting (HoG/Dense SIFT AAMs as reference) | landmark fitting · LFPW | fitting error: 0.0398 vs 0.0372 (HoG AAMs), 0.035 (Dense SIFT AAMs) | Mixed |
Closest papers by meaning
- Fast Algorithms for Fitting Active Appearance Models to Unconstrained Images (IJCV 2016 · similarity 0.89)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.85)
- Longitudinal Face Modeling via Temporal Deep Restricted Boltzmann Machines (CVPR 2016 · similarity 0.85)
- Temporal Non-Volume Preserving Approach to Facial Age-Progression and Age-Invariant Face Recognition (ICCV 2017 · similarity 0.84)
- Face Super-resolution Guided by Facial Component Heatmaps (ECCV 2018 · similarity 0.84)
- Face Normals "In-The-Wild" Using Fully Convolutional Networks (CVPR 2017 · similarity 0.84)
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.
