Repository for video classification projects using TensorFlow. Video classification involves the categorization of video content into predefined classes or categories based on the actions, events, or objects present in the video. It plays a crucial role in various applications such as surveillance, sports analysis, medical imaging, and entertainment.
UCF101 - Action Recognition Data Set
UCF101 is an action recognition dataset of realistic action videos, collected from YouTube, containing 101 action categories. With 13320 videos from 101 action categories, UCF101 offers diversity in actions and challenges due to variations in camera motion, object appearance, pose, scale, viewpoint, background clutter, illumination conditions, etc.
The action categories are divided into five types:
- Human-Object Interaction
- Body-Motion Only
- Human-Human Interaction
- Playing Musical Instruments
- Sports
For detailed categories, refer to the UCF101 dataset website.
UCF101 dataset is provided by the First International Workshop on Action Recognition with Large Number of Classes, ICCV'13.
For more information and published results, visit the UCF101 dataset website.