Cortically Inspired Architectures for Action Recognition in Movie Clips

Video data is becoming a much more prevalent medium for transmitting information. While object classification in still images has been a standard task for the computer vision community, the importance of having computers understand video sequences is beginning to be recognized. One task that has emerged from automated video understanding is action classification. In this task, the computer must correctly classify a video sequence which shows an action such as hugging or shaking hands. The original task was introduced in [4]. In their paper, they explored the correlation between action classification and scene classification. This relationship is not explored in our project as we are only concerned with action classification.