Pinterest Roll out Skin Tone Filter in Quest for a “More Inclusive Way to Search”


Pinterest has over the years morphed into a platform with rather specific uses which are fairly consistent across much of their 200 million-strong user base. It is arguably most often used as a source of inspiration for endeavours such as livening up your cooking or perfecting your hair and beauty routine, with a study conducted by Pinterest themselves showing that 70% of their users search the platform to discover and save various looks and styles.

Hair and beauty tips feature prominently on the site and many influential fashion bloggers maintain a presence on the platform, but with so much imagery flooding the platform it can be difficult to find exactly what you are looking for. Apparently, one oft-reported difficulty relates to finding makeup and beauty products that correspond to a specific person’s skin tone, as many brands now specialise their products in such a way to ensure a better match.

In an effort to help in this regard, Pinterest engineer Laksh Bhasin announced in a recent Medium post that the platform will begin rolling out the beta version of a new search feature which enables users to refine results according to a given skin tone range. This filter is only available on searchers relating to specific beauty terms for rather obvious reasons.

With any feature such as this privacy was bound to be a concern; one which Pinterest have addressed by assuring their users that no information will be stored as a result of these searches. The slight downside of this is that you will have to select a skin tone range for each individual search, but it does mean that you are free from targeted ads and data-mining stemming from your use of the feature.

So how exactly does this work? Detecting a skin tone within an image is problematic to begin with, as variables such as lighting, shadows, prominence and blurriness may all create inaccuracies. Human evaluation would of course be the most reliable solution, but by the company’s own admission there is simply too much content on the platform to make this a viable option. Instead, Pinterest turned their attention to a method that works at scale; namely machine learning.

Making use of a third-party Face AI library from ModiFace, a company specialising in augmented reality and machine learning for beauty applications, alongside deep neural networks, the development team were able to produce successive algorithms for skin tone detection. Pinterest then passed results through their own human evaluation platform, named Sophia, in order to gather additional data and further perfect the machine learning algorithm ahead of launch.

For now the feature is limited to just four skin tone ranges; however the company plans to expand these ranges moving forward.

“As new Pins are added to the system, we incrementally run the skin tone detection algorithm on just those new Pins, so we continue to increase our coverage of skin tone data and improve results,” wrote Mr Bhasin.

Post a Comment

[blogger]

Author Name

Free Gift

Free Gift
Get immediate access to our in depth video training on the click by click steps required to get your successful online business started today

Contact Form

Name

Email *

Message *

Powered by Blogger.