Case Study - optimizing AI for the edge
Deploying AI to low powered devices
Summary
New Gradient developed and deployed a lightweight deep neural network for object detection on low-powered edge devices, enabling live chicken monitoring for our client Omlet, a leading pet products company.
AI deployed to low powered devices
Omlet aimed to offer their customers a low-powered, battery-operated IP camera system that would allow chicken owners to remotely monitor their coops. The primary goal was to enable users to verify if their chickens had safely returned to the coop at night and receive mobile app notifications about chickens left in the run or when all chickens were safely inside. This system needed to operate effectively in scenarios without access to mains power, requiring the remote IP cameras to function on an extremely low power budget while still providing accurate monitoring capabilities.
New Gradient's Edge AI solution for Omlet demonstrates our expertise in developing and deploying AI models for resource-constrained environments. By successfully addressing the challenges of power limitation and the need for accurate, real-time monitoring, we've enabled Omlet to offer a unique and valuable product to their customers. This project showcases our ability to apply cutting-edge AI technologies to create practical solutions that enhance everyday life. The success of this Edge AI implementation opens up possibilities for similar applications in various IoT and smart home scenarios, particularly where power efficiency and real-time processing are critical factors.
The AI model's ability to perform complex object detection tasks with minimal energy consumption was crucial in making this product viable for Omlet's customers. By optimizing the neural network for edge deployment, we ensured that the system could operate reliably in various environmental conditions, providing chicken owners with peace of mind about their flock's safety.
To meet Omlet's requirements, we developed a highly efficient AI solution. We trained and deployed a lightweight neural network specifically designed for low-powered edge AI devices. Our network achieves impressive object detection results while operating within an exceptionally small power budget. This innovative approach allows the system to provide users with accurate chicken counts and deliver live notifications from their monitoring cameras, all while running on battery power.
