aldrinjenson / anomaly-detection

Anomaly-Detection using LSTM based network

Code for Computer Science Final year project, titled: Multi-Layered Architecture for Anomaly Detection in Surveillance Networks

Abstract

Due to limited performance of manual monitoring, law enforcement agencies are having difficulty in capturing or preventing anomalous incidents. An optimized and feature-based, intelligent, anomaly detection framework which operates in Convolutional Neural Networks(CNN) with bi-directional Long-Short-Term Memory(LSTM) and Resnet50 Models with reduced time complexity can be used to make it more efficient and accurate. Horizontal scaling can be used to ensure that requests are handled efficiently in edge nodes. Thus, the project aims to create a multi-layered system for processing multiple live feeds on a cloud computing platform while ensuring easy scalability without compromising on performance.

Contributors

• Adithya Anilkumar (LSTM Model)
• Aldrin Jenson (Backend, Classifier Model, Ingestion server, DB)
• Gouri Hariharan (Front End)
• Nayana Vinod (Front End, DB, Auth)

Project Report

Check: Report

References

• http://www.svcl.ucsd.edu/projects/anomaly/dataset.htm

Special Thanks to

• Dr. Preetha Theresa Joy - HOD: CS department at Model Engineering College
• Mr. Sreekumar K : Assistant Professor - CS Department MEC