Anomaly Detection Using CNN with I3D Feature Extraction

Anomaly Detection Using CNN with I3D Feature Extraction

A.Nandini1 , Ashifa.K1 , B.Jyoshnapriya1 K.R.Jayavelu2,J.Jayaprakash2,G.Victo Sudha George2

1 IVYear B.Tech, Dept of CSE,2 Professor,Dept of CSE

1nandininaidu450@gmail.com,1ashifabasheer19@gmail.com,1jyoshnapriya250101@gmail.com

1,2 DR. M.G.R EDUCATIONAL AND RESEARCH INSTITUTE,Maduravoyal, Chennai-95, Tamil Nadu,India

Abstract— Anomaly detection is a critical task in various fields such as surveillance, healthcare, and industrial monitoring, aiming to identify patterns that deviate significantly from normal behavior.Video anomaly detection is inherently difficult due to visual complexity and variability. This work proposes a unique anomaly detection technique leveraging Convolutional Neural Networks (CNN) with Inflated 3D Convolutional Networks (I3D) for feature extraction. This involves training the CNN on a large dataset to learn normal behavior, enabling it to identify anomalies by recognizing deviations from learned patterns. Furthermore, our approach exhibits promising results in detecting various types of anomalies, including sudden changes, abnormal trajectories, and rare events. Upon detection of such activity, mail(notification) can be raised concerned people who can take immediate action.This research contributes a significant advancement in the field of anomaly detection, and holds potential for applications in surveillance, security, and industrial monitoring systems.

Keywords—Anomaly detection,I3D(Inflated3D) feature extraction,Convolutional neural network, Spatio-Temporal Features,Normal and abnormal event detection.