Object and Human Action Recognition From Video Using Deep Learning Models

Diagnosing Suspects by Analyzing Human Behavior to Prevent Crime by Using Deep and Machine Learning

“…We can see from the Figs. 5 and 6 graphs that the model's accuracy has become more linear after 90% accuracy, indicating that accuracy is between 90% to can conclude that it has also reached a point where it will vary substantially.…”

“…Human action reveals details about a person's identity, personality, and emotional functioning. Human Activity Recognition (HAR) is a critical element of suspect detection and it's aim to identify activities based on inspections of subjects' activity and ambient variables [1][2][3][4][5]. Suspects can be classi ed in a variety of ways, depending on factors such as criminal behavior, crime scene trends, and motive [6][7].…”

“…Machine and deep learning algorithms can determine if a crime scene is ordered and premeditated or unorganized and unplanned [8][9]. Using a computer vision approach [10,12], some relevant papers have been discussed about classifying objects and recognizing human action from a video [13][14][15] where used Long Short-Term Memory networks with convolutional neural network techniques. Another relevant paper [16,17], who gives an overview of various classi cation techniques for recognizing human activities from wearable inertial sensor data.…”

“…Here, author uses a deep neural network to identify the behaviors of a lot of persons in a video frame. Due to the high-quality video process [5,24], data travel time across the network, and the time it takes to access and present the feed on a screen, security cameras might lag or experience a delay [4,22]. However, to mitigate such a problem, we proposed a new architecture for our system that detects suspicious activity based on video footage in real-time and provides the most e cient and reliable performance.…”