Introduction:

The dataset is a collection of videos of different sports. All videos were collected from YouTube, so the videos contain diverse events, different venues, and various athletes. Continuous shots were extracted from the videos to generate sequences. Each extracted sequence is a continuous shot with no scene or camera changes.

 

Dataset:

The dataset can be downloaded here. The dataset consists of five classes of sports: long jump, high jump,vault (gymnastics), snatch (weightlifting), and javelin throw. Each class contains 30 distinct sequences, with each sequence a continuous shot of an actor performing the sport. Here are the snapshots of the videos:
• long jump
• high jump
• vault
• snatch
• javelin throw

 

File format

Each sequence is a seq file. To open the file, you will need Piotr's Image & Video Toolbox for Matlab (http://vision.ucsd.edu/~pdollar/toolbox/doc/).