The client is a leading provider of AI-based security solutions that utilize computer vision to enable codeless development and deployment of machine learning models and AI applications.
Their platform offers near real-time detection of safety gear, goggles, and facial recognition, among others, for critical infrastructure management in manufacturing and energy plants.
The client had developed a computer vision model as part of their AI-based security platform solution that was intended to help detect safety gear, goggles, and facial recognition in near real-time. The current solution, which was connected to five cameras, faced challenges such as lag time in rendering and processing, and scalability to input from more than 100cameras.
The client was facing two main challenges with their existing AI-based security platform solution. Firstly, their solution was unable to handle input from more than 5 cameras, resulting in lag time in rendering and processing. Secondly, the solution was not scalable enough to handle input from over 100 cameras, which limited the potential growth of the client's business. The client needed a solution that could address these challenges and enable them to scale up their platform while maintaining real-time processing capabilities.
Codvo Team undertook an in-depth analysis of the client’s AS-IS architecture and existing C# ML code to optimize and scale the platform. Our team optimized image processing by moving from bitmap to CV2 functions and storing images as objects rather than B64 strings. We also pushed frames to the platform rather than pulling to avoid RTSP-induced delay and manage data streams from multiple cameras using NATS. Additionally, we connected additional SSD drives for storage in an on-prem server and added GPU for on-prem model retraining and deployment.
The solution provided by Codvo.ai had a significant impact on the client's business.
Overall, our approach helped the client achieve their objectives, resulting in a more efficient, reliable, and profitable business.