Meta has released the second version of its Casual Conversations dataset, which is a dataset designed to help AI researchers evaluate their models for fairness and accuracy. The updated dataset contains 26,467 video monologues from 5,567 people in Brazil, India, Indonesia, Mexico, Vietnam, the Philippines, and the United States, and contains self-provided demographic information, such as age, gender, and disability status, and annotations on participants’ voice timbre, apparent skin tone, and activity or recording setup.
Image credit: Flickr user Focal Foto