Google Photos Has This Embarrasing Facial Recognition Error


Announced and rolled out shortly after their developer conference – Google I/O 15Google Photos is a backup service by the software giant that seeks to solve the issue of storage, sorting and surfacing of people’s digital photos.

I got on to the service immediately after it was rolled out and highly recommended it to anyone who could listen (offline) and all over Social Media.

One of the more touted features of this service is the ability to automatically recognize faces and group them together. Made possible by Google’s Machine Learning, this feature has the ability to even recognize and match a baby’s face and map out growth in a way humans even found hard or impossible as demonstrated on stage.

This – Google Photo’s facial recognition – has not been rolled out world wide. A footnote on the Google Photos website indicates that “[The facial recognition] feature isn’t available in all countries,” This is likely due to privacy laws, but I found a simple workaround that enabled the face detection in my Google Photos.

When I tried the steps listed above, I was able to get facial recognition on my pictures and was quick to note that it works very well on Caucasian faces and had a lot of mismatches with African ones and especially darker ones or ones in low lighting.

While appreciative of the fact that Machine Learning is hard, today I came across an embarrassing instance of Google Photo’s mismatch of an African face. Twitter user ‘jackyalcine’ shared a screenshot of his photos app where a friend is grouped under the tag ‘gorillas’. A search of the term ‘gorillas’ on the his photos also returned picture results of the same person.

We are hoping Google Machine Learning gets better with Afro faces to avoid more situations like this.


Yonatan Zunger, the Chief Architect of Social at Google immediately got in touch with @jackyalcine on Twitter and according to him, the team behind Google Photos was hard at work and releasing a fix on the same, something they are doing at the moment.