Market for photos of masked people to train biometric algorithms surging

One of the largest stock photo platforms in Japan is building a database of images of individuals wearing masks to train facial recognition algorithms, The Asahi Shimbun reports.

Named Pixta, the Tokyo-based company reportedly started training its face biometric systems in 2018.

Fast forward to June 2021, Pixta started offering datasets each containing 1,000 facial photos of Japanese wearing masks. In the following months, requests for such datasets increased threefold, signaling a definite market demand amidst the pandemic and related mask-wearing trends.

“We began receiving more inquiries for photos of people wearing masks, so we started offering them as a dataset,” Sayaka Fukumoto, a Pixta official in charge of customer services, told The Asahi Shimbun.

“We hope the service will help clients save time and energy collecting those photos for their research.”

Today, driven by this data, Pixta is now trying to expand its datasets further to help increase accuracy in identification rates for this specific demographic.

The price for a dataset of images edited for machine learning currently costs 165,000 yen (roughly $1,200) each, according to Fukumoto. Packages of unedited ones, on the other hand, are traditionally offered for 99,000 yen (roughly $730) each. Datasets are sometimes purchased ten at a time.

The call for images to enhance Pixta’s face biometrics dataset comes months after Getty Images introduced its Enhanced Model Release form, designed to help users consent to their photos being used to build biometric datasets.

“We must recognize that the increased use of biometric data contained in imagery to train AI/ML applications requires the need to ensure that we have obtained the model’s permission to use their image and data in this manner, and Getty Images is at the forefront of addressing these very real concerns,” Paul Reinitz, director of advocacy and legal operations counsel at Getty said in March.

It is unclear at the time of writing whether Pixta intends to put in place any such privacy safeguards when building these databases.

Biometric Update has reached out to the company and this article will be updated to include any responses from Pixta.

Article Topics

accuracy  |  algorithms  |  biometric data  |  biometrics  |  dataset  |  face photo  |  facial recognition  |  mask detection  |  periocular biometrics  |  training