Asia Pacific University Library catalogue

A project submitted in partial fulfillment of the requirement of Asia Pacific University of Technology and Innovation for the Degree of B.Eng (Hons) in Electrical and Electronic Engineering (UC4F1811EEE).

The main aim of this project is to develop a video analytics system using neural networks to detect human action in live video stream. In this proposed method, a Graphical User Interface using MATLAB,was developed to demonstrate the proposed system. The performance of the developed proposed system is evaluated by testing the Average Precision Rate, Average Miss Rate, Network Latency Rate, Transmission Latency rate and Frame Drop Rate. The implemented system is observed to detect three different human actions wit a probable accuracy of 99% with the developed neural network. External UCF101 dataset as well as internally recorded videos used of neural network is trained with GTX 1060 GPU in MATLAB Software. The average precision for the detector is 0.7 and average miss rate is 0.5. The frame rate for the detector is 0.61 fps and 99% accuracy conditional detector has a frame rate of 0.14. The latency of camera transmission is 10ms at 5 meter and 4ms at 0 meter. The average alert transmission is 4 seconds. Finally, the system proved the concept of neural network for activity with precision of 70% in similar environment.