Case Study: STMicroelectronics achieves real-time low-power vision AI on microcontrollers with Focoos AI

STMicroelectronics deploys 2,000+ edge AI models with Focoos AI on STM32N6

STMicroelectronics faced the challenge of deploying modern computer vision models on their low-power edge microcontrollers, a process that was typically slow and manual. To make vision AI scalable on devices like the STM32N6, they needed a fully automated, hardware-aware, and efficient solution that did not require manual tuning or machine learning expertise. They partnered with Focoos AI to address this.

Focoos AI implemented its ANYMA Neural Architecture Search engine to automatically generate, evaluate, and deploy thousands of neural network architectures optimized for the STM32N6 hardware. This approach removed the need for handcrafted designs and iterative tuning. The resulting models delivered real-time performance with higher accuracy and speed compared to existing solutions, successfully running natively on the microcontroller without any manual adjustments required.

STMIcroelectronics