People detection and notifications based on the Raspberry Pi + AI Camera (Sony IMX500)

This project is a proof of concept based on the Raspberry Pi AI camera, equipped with the Sony IMX500 imaging sensor, which allows running vision AI models even with very limited hardware (e.g. the project has been tested on a Raspberry Pi 3 model B).

For testing purposes, the application can also run in a Linux laptop with a standard webcam. In this configuration, it uses a light-weight, pre-trained SSD MobileNetV2 model on CPU.

Disclaimer

This is an experiment, for fun. Expect rough edges and opinionated choices. For example, the only notification target supported in this version is Slack just because it suited my needs. In any case, it's trivial to add support for other sensors, other notification targets (e.g. s3 bucket, email, etc.)

Requirements

• Python 3.10 or higher
• Raspberry Pi hardware:
  • Raspberry Pi (ideally 4 or 5, I've tested successfully with 3B)
  • IMX500 AI camera

Setup

1. Hardware

To set up the Raspberry Pi AI camera, refer to the official documentation.

For testing, any Linux system with a webcam should do it (including a Raspberry Pi, although expect it to be slow if the model runs on the CPU)

2. Prepare the environment

pip install -r requirements.txt

3. Run the application

To report detections to console, run the command below, selecting webcam or imx500 as the desired sensor:

python3 aimonitor.py --sensor=[webcam|imx500] --console

To report to a Slack channel

python3 aimonitor.py --sensor=[webcam|imx500] --slack=<SLACK_CHANNEL_ID>

Note that the Slack notifier expects a bot token in a SLACK_TOKEN environment variable.