Making buildings and offices smarter means placing people at the heart of the environments we live and work in. This paradigm shift requires a particular attention by our surrounding electronics on the sensing intelligence. In this context, the people counting functionality becomes of paramount importance to adjust in real-time heater, HVAC or occupancy management. It results in more comfort for occupants and higher energy management efficiency.
Approach
– This prototype can count in real-time and with a high level of accuracy the restaurant’s attendance, while running on the standard STM32 microcontroller.
– This is achieved thanks to the artificial intelligence algorithm embedded on the STM32 microcontroller and the use of a thermal infrared technology.
Sensor
Vision: Camera module bundle (reference: B-CAMS-OMV)
Data
Data format:
Single class: people
RGB color images
Results
Model: YOLO Low Complexity quantized neural network
Input size: 240x240x3
Memory footprint:
277 KB Flash for weights
233 KB RAM for activations
Accuracy: 55.88% Average Precision using a 50% IoU against the PASCAL VOC test dataset
Performance on STM32H747* (High-perf) @ 400 MHz
Inference time: 371 ms
Frame rate: 2.7 fps
* As measured with STM32CubeAI 7.1.0 in FP-AI-VISION1 3.1.0
