Image
Compliance Assurance System

Overview

The Compliance Assurance System is a microservices-based architecture for pharma manufacturing plant. The system's primary objective is to detect compliance violations, such as uniform and safety gear adherence. Utilizing TensorFlow's object detection model, the solution processes live CCTV feeds and detects violations. Two significant challenges were addressed: the client's reluctance to directly send the stream to the AI server and the limited computational capacity of the edge device handling the stream.

Technologies and Frameworks

Industry

Automation

Project timeline

8 Months

No of Developer

5

Challenges and solution

Limited computation on edge device.

  • Implemented a microservices architecture that collected frames on the edge, storing them locally. Periodically, the frames were sent in batches to Amazon S3, reducing the computational load on the edge device and improving system throughput by 50%.

Restrict sending live stream outside network

  • Developed a decoupled approach where frames were collected on the edge device, stored locally, and then transferred to S3 in batches. The AI engine, hosted on an EKS Kubernetes cluster, fetched frames from S3 for object detection, ensuring compliance with client preferences.

Ensuring load balancing on AI Engine

  • Orchestrated the AI engine deployment using Kubernetes on an EKS cluster, ensuring scalability, resource optimization, and efficient management of the object detection model.

Efficiently storing and retrieving frames for AI processing.

  • Utilized Amazon S3 for secure and scalable storage, providing a centralized repository for frames that facilitated efficient retrieval and processing by the AI engine.

Screenshots

Previous Next