This talk will discuss low-level video processing. We will discuss both features and feature descriptors that have been proposed in the last two decades in video analysis. To ground the discussion, activity recognition will be used, but the emphasis is on the features directly. Example features to be discussed are space-time interest points and trajectories. Example descriptors to be discussed are 3D histograms of oriented gradients and motion boundary histograms.
Back to Graduate Summer School: Computer Vision