Motion detection and monitoring for Edge impulse

Story

The idea is to connect many sensors and when motion is detected, use the camera and detect different objects, for example, follow animals or monitor movement in the office.

Data collection

To stm32 connection of 4 motion sensors from which data is collected.

Model Training

I used raw data processing block and left default training model as my data is very simple I got good result.

Model FOMO for ESP32 CAM

For ESP32 CAM, photos of people from the Internet were collected, the image size is 64x64, after training, an accuracy of 80% was obtained, I decided to recognize people, but you can train anything.

Assembly and testing

I used my board with stm32f411 but any board can be used

camera module esp32 cam but you can use another as well as 4 motion sensors and one servo motor.

Testing

When motion is detected from one of the sides, the maximum recognition class will be sent to the esp32 camera via usart, and the FOMO model for object recognition will be launched there and the camera will be directed in the direction of movement

Links for projects

Motion monitoring - https://studio.edgeimpulse.com/public/158069/latest

Person detection - https://studio.edgeimpulse.com/public/121370/latest