Active shape models—their training and application
Computer Vision and Image Understanding
Neural Network-Based Face Detection
IEEE Transactions on Pattern Analysis and Machine Intelligence
Robust Real-Time Face Detection
International Journal of Computer Vision
Facial Attractiveness: Beauty and the Machine
Neural Computation
Active Shape Models with Invariant Optimal Features: Application to Facial Analysis
IEEE Transactions on Pattern Analysis and Machine Intelligence
Data-driven enhancement of facial attractiveness
ACM SIGGRAPH 2008 papers
Robust coding schemes for indexing and retrieval from large face databases
IEEE Transactions on Image Processing
A survey of perception and computation of human beauty
J-HGBU '11 Proceedings of the 2011 joint ACM workshop on Human gesture and behavior understanding
"Wow! you are so beautiful today!"
Proceedings of the 21st ACM international conference on Multimedia
| Hi-index | 0.00 |
This paper presents statistical analyses for facial beauty study. A large-scale database was built, containing 23412 frontal face images, 875 of them are marked as beautiful. We focus on the geometric feature defined by a set of landmarks on faces. A normalization approach is proposed to filter out the non-shape variations – translation, rotation, and scale. The normalized features are then mapped to its tangent space, in which we conduct statistical analyses: Hotelling's T2 test is applied for testing whether female and male mean faces have significant difference; Principal Component Analysis (PCA) is applied to summarize the main modes of shape variation and do dimension reduction; A criterion based on the Kullback-Leibler (KL) divergence is proposed to evaluate different hypotheses and models. The KL divergence measures the distribution difference between the beautiful group and the whole population. The results show that male and female faces come from different Gaussian distributions, but the two distributions overlap each other severely. By measuring the KL divergence, it shows that multivariate Gaussian model embodies much more beauty related information than the averageness hypothesis and the symmetry hypothesis. We hope the large-scale database and the proposed evaluation methods can serve as a benchmark for further studies.