Microsoft this morning announced that they have updated their facial recognition tech to better recognize gender across skin tones. A total of three major updates help address concerns that facial recognition tech more accurately recognized the gender of people with lighter skin tones than darker skin tones.
The company says they have expanded and revised training and benchmark datasets, launched new data collection efforts and improved classifiers as part of these changes. This allows for Microsoft to reduce the recognition error rates for men and women with darker skin by up to 20 times, and women by nine times. Microsoft also added they were able to significantly reduce accuracy differences across the demographics.
This facial recognition technology is available via Azure Cognitive Services, and the Azure team worked with experts on bias and fairness to improve a gender classifier system for getting better results for all skin tones. “We talked about data collection efforts to diversify the training data. We talked about different strategies to internally test our systems before we deploy them,” explained Hanna Wallach, a senior researcher in Microsoft’s New York research lab.
Further reading: AI, Artifical Intelligence, Microsoft