A Compositional and Dynamic Model for Face Aging

Abstract

In this paper we present a compositional dynamic model for face aging. By augmenting the high resolution hierarchic face model studied in Xu et al [40], [41] with aging and hair features, this compositional model represents all face images by a hierarchical And-Or graph [5]. The And nodes decompose face into parts and primitives at three levels from coarse to fine. The first level describes the low resolution face image including the hair-style and face appearance. The second level refines the first level by modeling shape, intensity and other variability of facial components. The third level accounts for the skin marks and wrinkles in different facial zones. Correspondingly three prominent aspects of aging related changes are integrated in this model: global appearance changes in hair style and face shape, deformations and aging effects of facial components, and wrinkles in various facial zones. Among the nodes at the same level, spatial relationships and constraints are modeled to force the validness of the configurations. At each level, the Or nodes describe the alternative sub-configurations, which model the diversity of human face across all ages.After choosing alternative of each Or nodes, the And-Or graph turns into a parse graph, which represents a specific instance of human face. Then face aging is modeled as a dynamic Markov process on this graph representation. The evolution of a graph includes not only the changes of its topology representing the emergence of features related to a new age but also the changes of its nodes attributes accounting for the extent of the aging status. The parameters of the dynamic model are learned from a large annotated dataset in a supervised way. Based on this model, we present an algorithm of synthesizing aged face images from a young face image: given an input image, we firstly compute its parse graph representation, and then sample the graph parameters over various age groups according to the learned dynamic model, including abrupt changes and continuous changes. Our sampling process accounts for the uncertainty in face aging by generating multiple plausible aged faces according to the stochastic model. Finally we generate new face images with the sampled parse graphs. We study two criteria to evaluate the synthesized aging results and conduct two human experiments on the aged image sequences synthesized by our algorithm. The results validate the performance of our model and aging algorithm.

Semi-Supervised FaceGAN for Face-Age Progression and Regression with Synthesized Paired Images

“…Before deep learning, the two main approaches for age progression were physical model approaches and prototype approaches. The physical model approaches focused on the change of physical factors (e.g., hair, wrinkles, mouth) over time [14,15]. Those approaches were complicated and required a large amount of paired data.…”

Gait-based age progression/regression: a baseline and performance evaluation by age group classification and cross-age gait identification

“…We evaluated the simulation results of the proposed method on the OU-ISIR Gait Database, Large Population Dataset with Age (OULP-Age) [46], which is the world's largest gait database with age information. Similar to face-based age progression/regression works [40,41], we conducted both subjective (human perception) experiments and objective (machine perception) experiments as quantitative measurements by implementing age group classification and cross-age gait identification to validate both the quality of aging patterns and preservation of identity for the simulation results.…”

“…Currently, extensive studies on face-based age progression have been conducted, with approaches mainly classified into two groups: physical model-based and prototype-based. Physical model-based methods correlate biological facial aging patterns with human age using complex models, such as the statistical face model [18] and-or graph [40,41], concatenational graph evolution aging model [39], and craniofacial growth model [33]. By contrast, a limited number of works on facial age regression [9,18,21] have been based on physical models by simply removing textures from facial surfaces.…”

“…Additionally, based on the assumption about a uniform transition between adjacent age groups, we can present the simulation of any target age by adjusting the morphing ratio during the transformation, which helps to improve the effectiveness for cross-age gait recognition in surveillance and the attractiveness for applications, such as entertainment, with a complete year-by-year aging/reverse aging process. By contrast, many face-based age progression/regression approaches [38,40,41,47]…”

“…To reduce the appearance difference caused by the age gap between the gallery and probe, modeling the gait aging process is one of the possible solutions. More specifically, gait aging modeling includes a simulation with natural aging and reverse aging effects, that is, age progression (i.e., prediction of the future gait) and age regression (i.e., estimation of the previous gait), which is similar to the definition of aging and reverse aging effects considered in existing studies on face analysis [18,[38][39][40][41]43,47]. However, in real scenarios, such as surveillance, facial images may not work well because of low image resolution or even occlusion by a face mask.…”