FaceRecognition_IOT

Face recognition using Deep Learning running on a Jetson Nano, which receives streaming video data from a Raspberry Pi via MQTT.

Setup

• 1 Raspberry PI (better a 4 or 3B+) will be the MQTT Broker, there you need to install Raspberrypi OS and run on a terminal "sudo apt-get install -y mosquitto mosquitto-clients".

• On the Raspberry PI (or even more than 1) which will stream the data, install the Raspberrypi OS Lite version and run on a terminal "sudo bash setup.sh".

• Setup the Jetson Nano by following the instructions provided here.

• If you want to see the data while it is streamed, you can either use your PC and run "pip3 install paho-mqtt opencv-contrib-python" or do the same on the Raspberry PI used as a MQTT Broker.

NOTE: the code used for training and inference on the Jetson Nano will be uploaded soon...

Usage

VideoStream.py

usage: VideoStream.py [-h] [--broker_ip BROKER_IP] [--topic TOPIC] [--max_fps MAX_FPS] [-f] [--resize RESIZE]

MQTTVideoStream, program created by David Forino AI Solutions (https://davidforino-aisolutions.com). Stream video from your camera via MQTT!

optional arguments:

• -h, --help show this help message and exit
• --broker_ip BROKER_IP IP address of your MQTT broker
• --topic TOPIC On which topic to stream the video
• --max_fps MAX_FPS Maximum FPS desired value
• -f, --face_detection Apply face detection and send only the image cropped to the face. If no face is detected, sends nothing. If multiple faces are detected, sends the biggest one only.
• --resize RESIZE Resize image to <width,height>

ViewStream.py

usage: ViewStream.py [-h] [--broker_ip BROKER_IP] [--topic TOPIC] [--timeout TIMEOUT] [--save_dir SAVE_DIR]

MQTTViewStream, program created by David Forino AI Solutions (https://davidforino-aisolutions.com). Visualize and save images stream via MQTT!

optional arguments:

• -h, --help show this help message and exit
• --broker_ip BROKER_IP IP address of your MQTT broker
• --topic TOPIC On which topic to stream the video
• --timeout TIMEOUT Timeout to close the live view window but keeps the connection
• --save_dir SAVE_DIR Directory where to store the images, if this argument is given, it won't show any image