: UCF101: A Dataset of 101 Human Action Classes From Videos in the Wild
: Actions are divided into five types: Human-Object Interaction, Body-Motion Only, Human-Human Interaction, Playing Musical Instruments, and Sports. Common Use Cases
: Unlike earlier datasets filmed in controlled labs, these videos are collected from YouTube and contain "in the wild" challenges like poor lighting, camera shake, and cluttered backgrounds.
: Extracting spatial-temporal features using models like I3D or C3D.
: Using pre-split training/testing sets defined in the paper to benchmark a new AI model's accuracy.
