Real-time action localization and prediction / Ahmed Ali Hammam ; Supervised Aboul Ella Hassanien , Mona Solyman

Thesis (M.Sc.) - Cairo University - Faculty of Computers and Artificial Intelligence - Department of Information Technology

The humans abilities enables them to analyze, interact, make decisions, plan, and memorize related action with objects in their surroundings. Among these skills, the visual system and visual information plays the main part in understanding the relationship of object, the actions and the dynamics of the real world. Thus, amazing and collaborate system give us the ability of recognizing objects, identify faces and actions, as well as expected events in advance. The main goal of computer vision and deep learning is to help computers in learning to stimulate human ability and perception. Every day people capture a large collection of videos and streaming using their devices, share them on social media. Millions of security, surveillance cameras around the world record billions of hours of video. With this large influx of Big Data, it has become impractical for humans to view and distill useful information from the collected data. Therefore computer vision algorithms for human detection, tracking, segmentation, action recognition, video retrieval, abnormal event detection and video summarization are becoming increasingly important. Detecting the locations of multiple actions in videos and classifying them in real-time are challenging problem termed 3action localization and prediction3 problem. Action localization has a wide variety of applications from monitoring and security in surveillance videos such as intelligent surveillance systems, auto-monitoring of patients system that can tell and recognize the difference between somebody sleeping or fainting, health care assistance, gaming, human-computer interaction(HCI), robotic, vehicle accident avoidance, criminal activity analysis, and sign language detection, video search, action retrieval and video indexing

Issued also as CD