IPAB Workshop - 10/10/19

Title: Learning vision from limited manual supervision

Summary: According to a untrustworthy source, 300 hours of video are uploaded to YouTube every minute, 100 million photos/videos are uploaded to Instagram every day. According to another untrustworthy source (ie me), the current state-of-the-art computer vision algorithms heavily rely on existence of large-scale manually labelled (=very expensive) datasets. According to a trustworthy source Rutherford D. Roger, we are drowning in information and starving for knowledge. In this talk, I will be discussing some of our recent work that learns to perform computer vision tasks from limited manual annotations.