Google has open-sourced the 10-point Monk Skin Tone scale (MST) for researchers and developers of unbiased AI. In the coming months the company plans to roll it out across a range of its products.
During Google’s annual developer conference I/O 2022, the company demonstrated how applying the scale can improve Google Search. For example, when searching for cosmetics, users can refine by skin tone and receive more relevant images.
The company is working on a standardized method of tagging web content, enabling authors and brands to tag content with attributes such as skin tone, hair colour and texture.
The MST scale will also appear in Google Photos. The company will update the AI-powered photo editing feature Real Tone for Pixel phones introduced in 2021. According to the developers, the MST scale will improve rendering of faces across all skin tones.
New Real Tone filters will appear in Google Photos on all available platforms in the coming weeks.
MST was developed in collaboration with Harvard professor Ellis Monk. It is intended to cover a broader range of skin tones.
The company also said that the scale will help create more representative datasets for training and testing AI models to address bias.
During I/O 2022, Google unveiled a number of other products and features for its AI-powered services.
In April, Google Lens added simultaneous image-and-text search.
In October 2021, the tech giant unveiled the Pixel smartphones with a Tensor Processing Unit for machine learning.
Subscribe to ForkLog AI news on Telegram: ForkLog AI — all the news from the world of AI!