Improving Borderline Adulthood Facial Age Estimation through Ensemble Learning

Authors: Anda, Felix; Lillis, David; Kanta, Aikaterini; Becker, Brett; Bou-Harb, Elias; Le-Khac, Nhien-An and Scanlon, Mark

Publication Date: August 2019

Publication Name: The 8th International Workshop on Cyber Crime (IWCC), held at the 14th International Conference on Availability, Reliability and Security (ARES)

Abstract:

Achieving high performance for facial age estimation with subjects in the borderline between adulthood and non-adulthood has always been a challenge. Several studies have used different approaches from the age of a baby to an elder adult and different datasets to measure the mean absolute error (MAE) that has been oscillating between 1.47 to 8 years. The weakness of the algorithms specifically in the borderline has been a motivation for this paper. In our approach, we have developed an ensemble technique that improves the accuracy of underage estimation in conjunction with our deep learning model (DS13K) that has been fine-tuned on the Deep Expectation (DEX) model. We have achieved an accuracy of 68\% for the age group 16 to 17 years old, which is 4 times better than the DEX accuracy for such age range. And we also present an evaluation of existing cloud-based and offline facial age prediction services such as Amazon Rekognition, Microsoft Azure Cognitive Services, How-Old.net and DEX.

Download:

Download Paper as PDF

BibTeX Entry:

@inproceedings{anda2019borderlineadulthood,
author={Anda, Felix and Lillis, David and Kanta, Aikaterini and Becker, Brett and Bou-Harb, Elias and Le-Khac, Nhien-An and Scanlon, Mark},
title="{Improving Borderline Adulthood Facial Age Estimation through Ensemble Learning}",
booktitle="{The 8th International Workshop on Cyber Crime (IWCC), held at the 14th International Conference on Availability, Reliability and Security (ARES)}",
series = {ARES '19},
year="2019",
month="08",
location={Canterbury, UK},
publisher={ACM},
address = {New York, NY, USA},
abstract="Achieving high performance for facial age estimation with subjects in the borderline between adulthood and non-adulthood has always been a challenge. Several studies have used different approaches from the age of a baby to an elder adult and different datasets to measure the mean absolute error (MAE) that has been oscillating between 1.47 to 8 years. The weakness of the algorithms specifically in the borderline has been a motivation for this paper. In our approach, we have developed an ensemble technique that improves the accuracy of underage estimation in conjunction with our deep learning model (DS13K) that has been fine-tuned on the Deep Expectation (DEX) model. We have achieved an accuracy of 68\% for the age group 16 to 17 years old, which is 4 times better than the DEX accuracy for such age range. And we also present an evaluation of existing cloud-based and offline facial age prediction services such as Amazon Rekognition, Microsoft Azure Cognitive Services, How-Old.net and DEX."
}