[Paper Review] Modelling of Facial Aging and Kinship: A Survey

Markos Georgopoulos, Yannis Panagakis+1 more|arXiv (Cornell University)|Feb 13, 2018

Face recognition and analysis191 references3 citations

TL;DR

This survey provides a comprehensive review of computational facial modeling for aging and kinship, analyzing datasets, facial representations (geometric, hand-crafted, learned), evaluation protocols, and key results. It identifies challenges and future directions for robust age-invariant and kinship-based facial analysis in real-world conditions.

ABSTRACT

Computational facial models that capture properties of facial cues related to aging and kinship increasingly attract the attention of the research community, enabling the development of reliable methods for age progression, age estimation, age-invariant facial characterization, and kinship verification from visual data. In this paper, we review recent advances in modelling of facial aging and kinship. In particular, we provide an up-to date, complete list of available annotated datasets and an in-depth analysis of geometric, hand-crafted, and learned facial representations that are used for facial aging and kinship characterization. Moreover, evaluation protocols and metrics are reviewed and notable experimental results for each surveyed task are analyzed. This survey allows us to identify challenges and discuss future research directions for the development of robust facial models in real-world conditions.

Motivation & Objective

• To provide a complete and up-to-date inventory of annotated datasets for facial aging and kinship research.
• To analyze and compare geometric, hand-crafted, and learned facial representations used in aging and kinship modeling.
• To review standardized evaluation protocols and metrics across age progression, age estimation, and kinship verification tasks.
• To synthesize notable experimental results and identify performance gaps and limitations in current methods.
• To highlight open challenges and suggest future research directions for robust facial modeling in real-world applications.

Proposed method

• Systematic review of recent literature on facial aging and kinship modeling from 2010 to present.
• Categorization and analysis of facial representations into geometric, hand-crafted, and deep-learned features.
• Compilation and critical evaluation of publicly available annotated datasets for aging and kinship tasks.
• Synthesis of evaluation protocols, including standard splits, metrics (e.g., accuracy, AUC), and benchmark results.
• Cross-task comparison of performance across age progression, age estimation, and kinship verification.
• Identification of methodological inconsistencies and gaps in reporting and evaluation practices.

Experimental results

Research questions

• RQ1 What are the most comprehensive and up-to-date annotated datasets available for facial aging and kinship modeling?
• RQ2 How do geometric, hand-crafted, and learned facial representations compare in performance for aging and kinship tasks?
• RQ3 What are the standard evaluation protocols and metrics used across age progression, age estimation, and kinship verification?
• RQ4 What are the leading performance results reported in the literature for each of the core tasks?
• RQ5 What persistent challenges and limitations hinder the development of robust facial models in real-world scenarios?

Key findings

• A comprehensive list of annotated datasets for facial aging and kinship is compiled, serving as a critical resource for future benchmarking.
• Learned representations, particularly deep neural networks, consistently outperform hand-crafted and geometric features in most aging and kinship tasks.
• Significant variability exists in evaluation protocols and metrics, limiting direct comparison across studies and highlighting a need for standardization.
• Despite progress, performance degrades substantially in real-world conditions, especially for extreme age differences and non-ideal imaging conditions.
• Kinship verification remains challenging, particularly for distant relatives and in the presence of large pose and lighting variations.
• Age estimation and progression models show improved accuracy but still struggle with long-term aging patterns beyond 20–30 years.