How does Svetha Venkatesh approach activity recognition?
My approach to activity recognition is deeply rooted in machine learning and feature engineering. We start by identifying salient features within video data that characterize specific actions. This could involve tracking skeletal movements, analyzing object interactions, or detecting spatio-temporal patterns. These extracted features are then fed into machine learning models, often designed for temporal data, to learn and classify different activities. The process is iterative, refining features and models based on experimental validation.
Ask Svetha Venkatesh the follow-up →