On this page you will find materials related to this research, such as lecture slides, as well as information about the dataset and how to obtain the code.
This project required the collection of a comprehensive dataset of stereo video sequences and corresponding depth sequences computed using the block-matching and semi-global block-matching stereo algorithms. As far as we know, this is the largest action dataset containing depth information. This is also the first dataset of its kind to include transition sequences for every pair of actions. 10 actors were recorded performing 20 transitions each, for a total of 200 transitions or 400 individual action sequences.
For information regarding how to obtain the dataset and/or code, please contact the Computer Vision Lab at UCF. For more information about the research, or for questions and comments about the code, please contact Stephan Boyer.
Here are the lecture slides for the project. No lecture slides were presented on weeks 1, 4, or 5.