Design Win

People counting application based on object detection Neural Network model on STM32

Solution Evaluation Kit
Core Products
CAD Resources
Tools & Software
Get Started

Solution Description

Space is a universally important commodity, especially in commercial and industrial spaces, where it is generally very scarce. People counting technology is central for adjusting layouts, thoroughfares, production lines, and so on… which is key for space and facility optimization.

There is an increased demand for smart building and smart factory solutions to help reduce operating costs, improve safety and identify opportunities to improve services. People flow detector and counter applications are especially useful for monitoring queueing and social distancing in today's post-pandemic world.

The challenge is to simplify software development and hardware design, especially for battery-based devices with processing, memory, and power constraints. Our solution addresses cost and design considerations by leveraging Artificial Intelligence on cost-effective, ultra-low-power STM32 microcontrollers.

How it works

The system captures digital video from a low-resolution image sensor and ensures efficient processing by a high-performance, dual-core STM32H7 microcontroller running an edge-based people counting application (FP-AI-VISION1) based on outputs generated by deep neural networks (DNN) optimized for STM32.

The MCU then displays the number of people counted, the captured video and the bounding boxes corresponding to people's positions in the image locally via a connected LCD screen or over a network.

The result is a cost-effective and low-power solution for people counting applications based on an object detection algorithm using a convolutional neural network (CNN) model topology optimized for STM32.

  • Key Product Benefits

    STM32H747XIH6 - High-performance, dual-core  microcontroller with DSP and DP-FPU instructions

    This microcontroller with Arm® Cortex®-M7 and Cortex®-M4 cores has the necessary peripherals to manage incoming digital video signals and the processing power and memory to ensure rapid and accurate people counting applications with minimal power consumption.

    FP-AI-VISION1 - STM32Cube function pack for Computer Vision artificial intelligence (AI) applications

    This software implements an advanced computer vision application using STM32 neural network libraries based on pre-trained models generated using the STM32CubeMX AI expansion pack (X-CUBE-AI).

  • All Features

    • Ready-to-use firmware featuring the implementation of based on an object detection algorithm using a convolutional neural network (CNN) model for real-time people counting applications
    • The edge processing approach ensures lower power consumption and latencies than centralized cloud solutions, with a focus on accuracy and personal privacy
    • Efficient low-power implementation based on the use of optimized AI models fitting in a microcontroller’s internal memory
    • Easy portability across different STM32 MCU series thanks to the STM32Cube ecosystem