Omniscient

Smart anomaly detection is a prime sought-after aspect, with multiple anomaly detection models having been effectively designed and implemented, however lacking the feature to provide this detection in real-time. The project deals with the real-time implementation of an anomaly detector to aid in ef

Project Title

Omniscient

Project Area of Specialization

Artificial Intelligence

Project Summary

Smart anomaly detection is a prime sought-after aspect, with multiple anomaly detection models having been effectively designed and implemented, however lacking the feature to provide this detection in real-time. The project deals with the real-time implementation of an anomaly detector to aid in efficient, effective, and timely surveillance. It takes based on the idea of bringing forth state-of-the-art edge device technology to help further the futuristic aspirations of not just regular but also smart cities of today.
With the use of methodical Deep Learning models, running on live-stream being collected from the CCTV cameras, the project aims to collect data and perform a detailed analysis to put forward coherent detection, in the form of an AI dashboard, and systematic reports. The proposed detection models chiefly pertain to the identification of robberies, violence, and vandalism, and aim to sound an alarm once abnormal behavior is detected to reduce the time lag in relevant action to the detected anomaly.
Security is a basic right for every individual, and is, therefore an aspect that needs to be tackled effectively, especially in the modern world today. With the increasing popularity of smart cities, gated communities, and restricted areas, the provision of heightened security is of utmost importance. It is due to these factors that we propose an edge-device-based solution to detect anomalies in real-time, through face recognition and face identification models.

Project Objectives

The objectives pertain to taking an input livestream from CCTV cameras and using that for the deployment of a real-time detection system, with efficient data analysis to perceive anomalies live, while being accessible via software and app.
Ensuring the use of lightweight Deep Learning models with edge computation technology, the project aims to attain systematic efficiency and performance while bringing forth a product that is futuristic and helpful to communities as a whole.

Project Implementation Method

The initiation of the project takes place with testing out the model to be deployed on pre-existing video footage. Following this process, the model will then be interlinked with live CCTV footage, with the model being trained on a GPU (minimum 1080 Ti), and then finally being tested and deployed on Jetson as an efficient edge-device product. CCTV will capture the live video stream and send it to jetson, jetson will process the video by the running the model(anomaly detection) on it and send the signals to the alarm and hit the APIs on the server which will update Omniverse(Our Software which handles the dashboard and more data for better training).
With the increasing popularity of smart cities, gated communities, and restricted areas, the provision of heightened security is of utmost importance. Due to these factors, we propose an edge-device-based solution to detect anomalies in real-time, through face recognition and face identification models. The structure of the current system makes use of two segregated entrance and exit gates allowing the CCTV cameras to face in the direction of the crowd. Both CCTV cameras are connected to Jetson Nano which will be used for running the models, collecting the data and logging it into the database. For this system, we assume two kinds of individuals in any gated community: registered and unregistered. A registered individual (RI) is one who is pre-identified as a resident or visitor of the community, while an unregistered individual (URI) is one who enters the community for the first time. For increased efficiency and added effectiveness, our future work will involve the use of Jetson Xavier in the system as well. The further implementation of the system sees the use of multiple edge device-mounted (Jetson Nano, Raspberry Pi etc) CCTV cameras planted in different areas of the society. Each of which will be running different models for various purposes and the data collected through these edge devices would be sent forward to Jetson Xavier, which will act as the brain of the system and will perform any and all further analytics on the collected data.

Benefits of the Project

The primary benefit of the project lies in its real-time detection which allows for effective surveillance, enhanced security, and a better-controlled environment. With the addition of embedded systems, the project implementation then opens doors to a blend of programs and equipment in a way that allows for an autonomous framework to run on the edge device.
With our proposed method, we would be able to provide a true real-time implementation of the anomaly detector, which takes into account ease of accessibility, and pertains to making it easy for the user to interact with, and adapt to. With the addition of the alarm to the system, once an anomaly is detected the user would be able to hear a ringing alarm, which would reduce the time delay for an appropriate action to be taken for the identified anomaly.
Another key benefit that we bring forward is that of our models being able to run offline as well, which when added to the batch running of multiple CCTVs, brings forth another layer of enhancement to the already existing surveillance systems of today.
Once deployed, this implementation would aid law enforcement agencies, and other stand-alone security service providers to obtain an additional layer of security for themselves and their clients; who range from the general public to high profile individuals; helping to create and nurture a secure environment.

Technical Details of Final Deliverable

The final deliverable is essentially a Jetson chip with real-time anomaly detection models integrated onto it. It can be further mounted onto any CCTV camera and perform detections while providing security to the users.
CCTV cameras are connected to Jetson Nano which runs the models, collects data and logs it into the database, thereby performing identification of all individuals in a given area.
Additionally, for enhanced effectiveness, a network created between multiple Jetson Nanos and Jetson Xavier, allows for communication between the individual chips. Thereby creating an end-to-end system that will improve security conditions of communities as a whole.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Security

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Industry, Innovation and Infrastructure, Responsible Consumption and Production

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com