Detect herons at the Pond

The Docker container monitors your pond.
If a heron is detected, an alarm is triggered via MQTT.
Inference is saved in the Image folder.

Required components:

• Raspberry Pi >= 4
• Installed Docker
• IP Camera (with zoom an advantage)
• MQTT Client, such as ioBroker, etc.

Optional:

• Coral Edge TPU USB Accelerator - ≈50 frames/s on Pi 4, without ≈7 frames/s

Create Folders:

• mkdir docker/tflite # Volume for the container
• mkdir coral # Add the repository files here

Edit heron.py

• Adjust your MQTT client
• Change the MQTT alarm ID
• Change the RTSP link of your camera
• Choose whether to use the Edge TPU accelerator or not

Build the Container:

cd coral
docker build -t "heron".
docker run -it --privileged --restart always \
    -e MTX_PROTOCOLS=tcp \
    -v /dev/bus/usb:/dev/bus/usb \
    -v /home/pi/docker/tflite:/tflite \
    heron /bin/bash