Automatic Genre Classification from Videos

In recent decades, there has been a huge growth in the amount of multimedia content stored in networked repositories. Many video hosting websites exists in today’s scenario such as youtube, metacafe and google video etc., where people are uploading and downloading their videos. At present, indexing and categorization of these videos is a tiresome job. Either the system asks the user to suggest tags for the videos which they upload or people are employed to tag the video manually. Manual tagging has been done based on the views of users, search terms, etc. In order to eliminate this problem, this paper proposes a model that automatically categorizes the videos based on their genres. The main aim of this work was to categorize the videos broadly on major domains such as sports, music and news using temporal, textural, motion, and color features. In sports, the videos have been classified further into cricket and football. The hierarchical SVM has been used for automatic training and selection of the genre of the video. A total of 350 videos from various Web sites have been used for training the classification system. This system achieves an overall average detection ratio up to 98 % while maintaining very low false detection rate of 2 %.