Explore visual attribute learning techniques in computer vision with Professor Kristen Grauman in this 31-minute lecture. Delve into the challenges of learning the right visual properties, methods for sharing between properties, and strategies to avoid learning the wrong thing. Examine success cases in category-specific attributes, transfer learning, and fine-grained attribute differences. Discover the concept of relative attributes and gain insights into local learning approaches for testing fine-grained attributes. Understand how competition for features impacts visual learning and explore the potential of local model overviews in advancing computer vision capabilities.
Kristen Grauman - Learning the Right Thing with Visual Attributes
Overview
Syllabus
Introduction
Learning the wrong thing
Learning the right thing
Sharing between properties
Competition for features
Success cases
Category specific attributes
Transfer learning
Results
Finegrain attribute differences
Relative attributes
Local learning approach
Local model overview
Testing finegrained attributes
Local Learning
Taught by
MITCBMM
