Loading...
Please wait, while we are loading the content...
Similar Documents
Face Recognition Vendor Test (FRVT) - Performance of Automated Age Estimation Algorithms
| Content Provider | Semantic Scholar |
|---|---|
| Author | Ngan, Mei Grother, Patrick J. |
| Copyright Year | 2014 |
| Abstract | Introduction Facial age estimation is an area of study new to the Face Recognition Vendor Test (FRVT) with Still Facial Images Track. While peripheral to automated face recognition, it has become a growing area of research, given its potential use in various applications. The motivation for age estimation systems has grown in the last few decades, given the rise of the digital age and the increase in human-computer interaction. Age-based access control and verification (e.g., age verification for alcohol/tobacco purchases), age estimation in crime and mass disaster investigation (e.g., age determination of unknown human bodies at a crime scene to help with victim identification), age-adaptive targeted marketing (e.g., displaying age-specific advertisements from digital signage), age-invariant person identification (e.g., identifying missing children), and age-based indexing of face images are potential applications of automated facial age estimation. NIST performed a large scale empirical evaluation of facial age estimation algorithms, with participation from five commercial providers and one university, using three large operational datasets comprised of facial images from visas and law enforcement mughots, leveraging a combined corpus of over 7 million images. NIST employed a lights-out, black-box testing methodology designed to model operational reality where software is shipped and used " as-is " without algorithmic training. Core age estimation accuracy was baselined over a large homogeneous population, then assessed demographically by age group, gender, and ethnicity. The impact of input-driven variations, namely image quality and number of image samples per subject was captured, and assessments of age-verification accuracy and estimation accuracy in children were documented. Core Accuracy and Speed: Age estimation accuracy depends strongly on the provider of the core technology. Broadly, there is a twofold difference between the most accurate and the least accurate algorithm in terms of the percentage of images correctly classified to within five years and mean absolute error (MAE) 1. Using the most accurate age estimation algorithm, (i.e., B31D from Cognitec), the chance of accurately estimating the age of a person within five years of their actual age over an ethnically-homogeneous database of 6 million images is 67%, with an MAE of 4.3 years. All algorithms can perform age estimation on a single image in less than 0.15 seconds with one server-class processor. The most accurate algorithm, on average, performs estimation in 0.125 seconds. The main dataset used for overall accuracy assessment is comprised of 6 million ethnically-homogeneous images. Although image collection was subject to the guidelines … |
| File Format | PDF HTM / HTML |
| DOI | 10.1002/https://dx.doi.org/10.6028/NIST.IR.7995 |
| Alternate Webpage(s) | http://www.nist.gov/customcf/get_pdf.cfm?pub_id=915238 |
| Alternate Webpage(s) | http://ws680.nist.gov/publication/get_pdf.cfm?pub_id=915238 |
| Alternate Webpage(s) | https://doi.org/10.1002/https%3A%2F%2Fdx.doi.org%2F10.6028%2FNIST.IR.7995 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
