AI on Edge Devices

Problem

The client wanted to track and profile individuals visiting their corporate properties while maintaining optimal performance within resource and computational constraints. They aimed to gather statistical data about occupation, assets, arrival experience, cleanliness. However, they encountered several pain points and obstacles. The first significant challenge was implementing complex computer vision models within their resource constraints devices.

Solution

To address this challenge, our team employed a series of strategies and tactics. They began by selecting and testing state-of-the-art models suitable for the real estate industry. These models come from various sides and some of the technologies used are Ultralytics YOLOv8, OpenCV, ResNet, TensorRT, Tensor Flow, and Pytorch. Leveraging edge devices offered on-site predictions, eliminating the need for sensitive data transmission over the network and minimizing network latency. The team focused on achieving a higher accuracy and frame rate, camera quality, accurate distance measurement, and optimal lighting conditions to ensure better performance compared to the client’s existing system.

Key metrics/technologies

The key metrics/technologies include higher frame rate and better accuracy and performance compared to the previous generation device, real-time processing and predictions on edge devices, and enhanced privacy and security compliance with GDPR regulations.

• Ultralytics YOLOv8,
• OpenCV,
• ResNet,
• TensorRT,
• Tensor Flow,
• Pytorch

Client

The client is a mid-sized US real estate company with over 500 employees. They have a well-established analytics system to track business performance and make data-driven decisions.