Google Open-Sources Skin Tone Research As An Attempt To Improve Skin Tone Evaluation In Machine Learning

This Article Is Based On The Research Article 'Improving skin tone representation across Google'. All Credit For This Research Goes To The Researchers πŸ‘πŸ‘πŸ‘

Please Don't Forget To Join Our ML Subreddit

As our world is so diverse, it’s becoming increasingly crucial that the vanguard of research is inclusive. Google has open-sourced its skin tone research to increase inclusiveness, intending to create a workplace that welcomes diversity. Until recently, the bulk of products and services catered to a specific skin tone. Google has been working with researchers to make its products and services more accessible to people with darker skin tones. The majority of this work is publicly available, and research so far has yielded the Monk Skin Tone Scale (MST), which will assist in the creation and growth of technology while reflecting a broader spectrum of skin tones.

Because seeing yourself in the results is crucial in locating the information that is relevant and useful. Google is rolling out changes to show a broader range of skin tones in image results for general queries regarding people or those where people appear in the results. In the future, Google plans to employ the MST Scale to recognize better and rank photographs, giving in a more extensive set of results so that everyone can find what they’re searching for.

Results can now be filtered according to a skin tone

The MST Scale will assist the tech sector in generating more representative datasets that will be used to train and assess AI models for fairness, resulting in features and products that perform better for everyone, regardless of skin tone. The scale will evaluate and enhance face detection algorithms in photos.

To begin, consumers will see this research in Google Search results and Google’s Photos app. usedThe MST scale is being used by Google to surface search results that are more inclusive of darker skin tones. Makeup-related searches, for example, will have a filter for adjusting for different skin tones so that users can discover the most relevant results. The MST scale is also being used by Google Photos to power a new set of “Real Tone effects.” Google intends to expand this study to include diverse hair textures and hues. Google hopes that by making the MST scale widely available to the more significant industry, companies would adopt it into their development processes and collaboratively enhance this area of AI.Β 



Aneesh Tickoo is a consulting intern at MarktechPost. He is currently pursuing his undergraduate degree in Data Science and Artificial Intelligence from the Indian Institute of Technology(IIT), Bhilai. He spends most of his time working on projects aimed at harnessing the power of machine learning. His research interest is image processing and is passionate about building solutions around it. He loves to connect with people and collaborate on interesting projects.

[Announcing Gretel Navigator]Β Create, edit, and augment tabular data with the first compound AI systemΒ trustedΒ by EY, Databricks, Google, and Microsoft