Machine Learning on Images with Noisy Human-centric Labels

When humans describe images, they have a reporting bias, in that the report only what they consider important. Thus, in addition to considering whether something is present in an image, one should consider whether it is also relevant to the image before labeling it. Ishan Misra joins us this week to discuss his recent paper Seeing through the Human Reporting Bias: Visual Classifiers from Noisy Human-Centric Labels which explores a novel architecture for learning to distinguish presence and relevance.

Guest

Ishan Misra: I work on computer vision and machine learning research specifically in generative AI and representation learning. I am a Director, Research Scientist in the TBD Labs research division at Meta's SuperIntelligence group. Previously I was at the GenAI group at Meta where I lead the research efforts on video generation models. I was the tech lead for Meta's Movie Gen project for foundation models in video generation, video editing, video personalization, and audio generation.