www.forensicsandsecurity.com Open in urlscan Pro
107.161.173.115  Public Scan

Form analysis
0 forms found in the DOM

Text Content

 * Home
   
 * People
 * Publications
 * Projects
 * Conferences
   * WCSF 2018
   * ECCWS 2017
   * WCSF 2017
     
 * Collaborations


VISAGE DATASET


 

The importance of large sample size, good quality annotated facial age datasets,
and the sharing thereof, with the research community is fundamental. Facial
photos datasets annotated with both age and gender are needed to train accurate
artificial intelligence models, especially for underage subjects. To that end,
we are creating a large human-annotated facial age dataset for the entire human
age range.

Dataset Curation Methodology
We gather creative commons and public domain photos with annotated ages from a
variety of online sources with indicative age metadata. We then process each of
these photos to detect faces, gender, and other associated metadata. Given the
level of error rates in automated facial age estimation and gender
identification, we subject each of these photos to human age and gender
verification. Each photo is voted on by three human assessors and if the
decisions on age and gender are unanimous, we add that photo to our dataset.

What's available right now?
We're offering an alpha version of the single subject dataset to digital
forensic researchers with 19,446 images in total across the 1-18 age range.

What human verified metadata is included for each image?

 * Age
 * Gender



What other metadata is included for each image?
We include the license, photo title, photo tags, photo's source URL, and
complete EXIF data for each photo (bar geolocation information). We also run
every photo through Microsoft Azure's Face API, which produces a range of
metadata including:

 * Facial Bounding Box
 * Emotion
 * Landmarks, e.g., pupil location, nose tip, lips, eyebrows, etc.
 * Pitch, yaw, and roll
 * Glasses (Yes/No, type)
 * Hair colour
 * Make-up

A sample of the Face API data included per photo is available here.



The collection, curation, and distribution of the VisAGe dataset has been
granted ethical approval from University College Dublin's Human Research Ethics
Committee with reference number LS-17-74. The photo metadata generation is
supported in part by a Microsoft Azure for Research grant.


FAQ

WHY DO I HAVE TO REGISTER FOR ACCESS?

One of the conditions for distribution of the dataset of our ethical approval
requires registration for access.

WHAT LICENCE IS THE DATASET RELEASED UNDER?

The dataset and associated metadata is released under the CC-BY 4.0 license.
Note: Not all photos permit modification or derivative works (see next
question). In terms of fulfilling the attribution requirement of the release
license, we kindly request that you cite our work in any
publications/presentations where you used our dataset and link back to this page
(https://forensicsandsecurity.com/visage.php).

Publications related to this dataset:

 * Anda, F., Le-Khac, N-A., and Scanlon, M., DeepUAge: Improving Underage Age
   Estimation Accuracy to Aid CSEM Investigation, Forensic Science
   International: Digital Investigation, ISSN 2666-2825, Elsevier, April 2020.
 * Anda, F., Becker, B., Lillis, D., Le-Khac, N-A., and Scanlon, M., Assessing
   the Influencing Factors on the Accuracy of Underage Facial Age Estimation,
   The 6th IEEE International Conference on Cyber Security and Protection of
   Digital Services (Cyber Security), Virtual Event, 15-19th June 2020.
 * Anda, F., Lillis, D., Kanta, A., Becker, B., Bou-Harb, E., Le-Khac, N-A.,
   Scanlon, M., Improving the Accuracy of Automated Facial Age Estimation to Aid
   CSEM Investigations, Digital Investigation, ISSN 1742-2876, Elsevier, April
   2019.
 * Anda, F., Lillis, D., Kanta, A., Becker, B., Bou-Harb, E., Le-Khac, N-A.,
   Scanlon, M., Improving Borderline Adulthood Facial Age Estimation through
   Ensemble Learnings, 8th International Workshop on Cyber Crime (IWCC), held at
   the 14th International Conference on Availability, Reliability and Security
   (ARES), Canterbury, UK, August 2019.
 * Anda, F., Lillis, D., Le-Khac, N-A., and Scanlon, M. Evaluating Automated
   Facial Age Estimation Techniques for Digital Forensics, 12th International
   Workshop on Systematic Approaches to Digital Forensics Engineering (SADFE),
   IEEE Security and Privacy Workshops, May 2018.



CAN I MODIFY THE PHOTOS OR CREATE DERIVATIVE WORKS?

While all photos contained within the dataset are licensed under public domain
or creative commons licenses, and permit redistribution, individual photos may
have restrictions regarding their modification or creation of derivative works.
The license for each image (as taken from the source of the image) is included
in the associated metadata.

DO YOU MODIFY THE PHOTOS?

We do not modify, scale or adjust the visual format of any of the photos. The
photos are shared in their original format and contain all associated metadata
from their original source with the sole exception of removing any embedded
geolocation data, as this information is beyond the scope of the dataset's
remit.

WHAT OTHER OPEN FACIAL AGE DATASETS ARE OUT THERE?

Other facial age datasets include IMDB, WIKI, FG-NET, MEDS, among others.
Combining and curating each of the aforementioned datasets together for age
ranges 0-19 results in the following total count of subjects:

Age Male Female Age Male Female 0 12 34 10 223 258 1 48 21 11 481 409 2 78 26 12
453 480 3 21 29 13 783 415 4 52 23 14 738 482 5 72 38 15 1027 565 6 99 51 16
1300 1949 7 136 65 17 1637 2484 8 129 155 18 1961 2856 9 174 190 19 2516 3427


DATASET ACCESS REQUEST

Loading...



SITEMAP

 * Home
 * People
   
 * Workshop


LINKS

 * UCD School of Computer Science
 * University College Dublin
   


LATEST NEWS

Congratulations to David Lillis and Mark Scanlon on winning Best Paper Award at
the 9th EAI International Conference on Digital Forensics & Cyber Crime in
Prague.

Read Paper

We're hosting the 3rd International Workshop on Cloud Security and Forensics in
Sydney, Australia in August 2017!

Learn More


GET IN TOUCH

Twitter @ForSecResearch Email WWW forensicsandsecurity.com
Address School of Computer Science,
University College Dublin
Ireland. Phone +353 1 7162930
 * © Mark Scanlon. All rights reserved
 * Design: HTML5 UP


HomePeoplePublicationsProjectsConferencesWCSF 2018ECCWS 2017WCSF
2017Collaborations