Paper Title
A Comprehensive Survey on Human Suspicious Activity Detection System for Defense Using Deep Learning

Abstract
In the fields of surveillance and defense, spotting questionable human behavior is essential to maintaining security and averting possible dangers. This research introduces a deep learning-this modelis use for detecting suspicious human activity in defence environment. Convolutional neural networks (cnns), lstm networks, and faster r-cnn are some of the sophisticated computer vision models that the suggested system uses to identify and categorize questionable activity in real time from video streams [1][2]. Robust action detection is combined with keyframe extraction and position estimation to enable the system to detect acts such as loitering, illicit access, and even weapon handling [3]. The framework is intended to process live video streams by utilizing a tensorflow/pytorch-based backend for model inference and opencv for frame analysis. Real-time warnings and remote monitoring are made possible by the system's integration of a firebase cloud-based architecture [4][5]. Keywords - Convolution Neural Network, R-CNN, Support Vector Machine, Acute Disease, Lstm.