Exposing.ai
Adience Benchmark Dataset
Search Datasets About
10 of about 2,284 people from the Adience age and gender estimation dataset. Faces are blurred to protect privacy. Visualization by Adam Harvey / Exposing.ai licensed under CC-BY-NC with original images licensed and attributed under Creative Commons CC-BY (attribution required, no commercial use).
10 of about 2,284 people from the Adience age and gender estimation dataset. Faces are blurred to protect privacy. Visualization by Adam Harvey / Exposing.ai licensed under CC-BY-NC with original images licensed and attributed under Creative Commons CC-BY (attribution required, no commercial use).

Adience Benchmark

The Adience Dataset was created in 2014 for age and gender estimation research and, according to the author, includes 26,580 photos with 2,284 identities. 2 Based on the downloads, there were slightly more than 2 face regions extracted per photo for a total of 10,805 unique photos from 168 Flickr user accounts. "The source for the photos in our set are Flickr.com albums, produced by automatic upload from iPhone 5 or later smartphones." 1 Though the dataset is not built explicitly for face recognition, it first appeared in Journal of Information Forensics and Security, hinting at potential applications to surveillance related technology. Age and gender can be used independent of identity recognition, and it can also contribute to overall performance of face recognition when used for feature clustering.

In fact, the authors the dataset note that the Adience dataset was created in "an effort to close the gap between the capabilities of age and gender estimation systems and face recognition systems." 1 After its publication in 2014, the Adience face dataset remains popular with international biometrics researchers and has been verified in over 100 research projects, including several research projects affiliated with a defense and commercial organization.

Based on an analysis of the Flickr metadata connected to the Adience dataset images, the majority of photos (over 9,000) are now marked as All Rights Reserved.

This page is still under development. Sorry for the delay.

Information Supply Chain

To help understand how Adience Benchmark Dataset has been used around the world by commercial, military, and academic organizations; existing publicly available research citing Adience Benchmark was collected, verified, and geocoded to show how AI training data has proliferated around the world. Click on the markers to reveal research projects at that location.

Citation data is collected using SemanticScholar.org then dataset usage verified and geolocated. Citations are used to provide an estimated overview of how and where images were used based on institutional affiliations. Thicker lines represent more citations. Please zoom in to see all institutions, as cities may have multiple points very close together.
Adience Benchmark Attributes
Dataset Name Adience
Dataset Name Full Adience Benchmark
Total Images 26,580
Videos 600
Identities 2,284
Initial Purpose Age and gender estimation
Year Published 2014
Download Size (GB) 1.34 GB
Dataset Website http://www.openu.ac.il/home/hassner/Adience/data.html
Photos from Flickr.com in Adience Benchmark
Total Flickr Photos 10,805
Total Flickr Users 139
Active on Flickr.com* 8,482
Inactive/removed on Flickr.com* 2,323
API Data Accessed October 2019
Included in YFC100M 744
Photos w/ Geo Data 1,435
Searchable on Exposing.ai 10,805

Adience Copyright Distribution

Years
Adience Creative Commons license distribution | Download Data (CSV) | Download Chart (SVG)

Adience Creative Commons License Distribution

Years
Adience Creative Commons license distribution | Download Data (CSV) | Download Chart (SVG)

Adience Image Upload Year Distribution

Years
Adience Creative Commons license distribution | Download Data (CSV) | Download Chart (SVG)

Top 10 Adience Image #Tags

Years
Top 10 image #tags used in Adience | Download Data (CSV) | Download Chart (SVG)

Top 10 Geocoded Cities Adience

Years
Top 10 cities for geocoded photos in Adience | Download Data (CSV) | Download Chart (SVG)

Citing This Work

If you reference or use any data from the Exposing.ai project, cite our original research as follows: 

@online{Exposing.ai,
  author = {Harvey, Adam. LaPlace, Jules.},
  title = {Exposing.ai},
  year = 2021,
  url = {https://exposing.ai},
  urldate = {2021-01-01}
}

If you reference or use any data from Adience cite the author's work:

@article{Eidinger2014AgeAG,
    author = "Eidinger, Eran and Enbar, Roee and Hassner, Tal",
    title = "Age and Gender Estimation of Unfiltered Faces",
    journal = "IEEE Transactions on Information Forensics and Security",
    year = "2014",
    volume = "9",
    pages = "2170-2179"
}

References