Analyzing camera feed using RedisAI, OpenCV and Grafana

Introduction

This project demonstrates how to analyze camera feed stored as Redis Streams using serverless engine RedisGears, RedisAI and display analyzed frames with metrics in Grafana.

Demo

Demo is available on demo.volkovlabs.io:

• Redis Overview dashboard
• Camera Feed dashboard

Requirements

• Docker to start Redis and Grafana.
• Python to run scripts.

Redis with OpenCV Docker image

This project provides Docker image with Redis, RedisTimeSeries, RedisGears, RedisAI and installed OpenCV for Python library.

docker run -p 6379:6379 --name=redis-opencv ghcr.io/redisgrafana/redis-opencv:latest

OpenCV requirements

Check that OpenCV with dependencies downloaded, installed and registered in the RedisGears requirements:

cluster.remote:6379> rg.pydumpreqs
1)  1) "GearReqVersion"
    2) (integer) 1
    3) "Name"
    4) "Pillow"
    5) "IsDownloaded"
    6) "yes"
    7) "IsInstalled"
    8) "yes"
    9) "CompiledOs"
   10) "linux-buster-x64"
   11) "Wheels"
   12) 1) "Pillow-8.2.0-cp37-cp37m-manylinux1_x86_64.whl"
2)  1) "GearReqVersion"
    2) (integer) 1
    3) "Name"
    4) "opencv-python"
    5) "IsDownloaded"
    6) "yes"
    7) "IsInstalled"
    8) "yes"
    9) "CompiledOs"
   10) "linux-buster-x64"
   11) "Wheels"
   12) 1) "numpy-1.20.3-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl"
       2) "opencv_python-4.5.2.52-cp37-cp37m-manylinux2014_x86_64.whl"
3)  1) "GearReqVersion"
    2) (integer) 1
    3) "Name"
    4) "numpy"
    5) "IsDownloaded"
    6) "yes"
    7) "IsInstalled"
    8) "yes"
    9) "CompiledOs"
   10) "linux-buster-x64"
   11) "Wheels"
   12) 1) "numpy-1.20.3-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl"

Import AI model and script

To load AI model and PyTorch script to the Redis database run:

cd ai/
cat tiny-yolo-voc.pb | redis-cli -h redis -x AI.MODELSTORE yolo:model TF CPU INPUTS 1 input OUTPUTS 1 output BLOB
cat ai-yolo-script.py| redis-cli -h redis -x AI.SCRIPTSET yolo:script CPU SOURCE

Start Grafana

Grafana can be started using Docker Compose or installed locally with Redis plug-ins for Grafana and Volkov Labs Image panel.

docker-compose pull
docker-compose up

When starting using Docker Compose, dashboard and plug-ins will be auto-provisioned and available in Grafana.

Register RedisGears script

Select Camera Processing dashboard and copy-paste gears-yolo.py script to RedisGears Script editor panel.

Click on the Run script button and you should see registered script StreamReader in the Registrations panel.

Start Camera

Copy script edge-camera.py and requirements.txt to IoT or any device with camera. Run script by specifying Redis URL, number of frames per second and rotate camera if required.

cd camera/
pip3 install -r requirements.txt
python3 edge-camera.py -u redis://redis:6379 --fps 6 --rotate-90-clockwise true

You should see output with Stream's Id and size of the jpeg file.

# python3 camera.py -u redis://red:6379 --fps 6 --rotate-90-clockwise true
Connected to Redis: ParseResult(scheme='redis', netloc='redis:6379', path='', params='', query='', fragment='')
id: b'1622145939609-0', size: 5479
id: b'1622145939769-0', size: 10006
id: b'1622145939928-0', size: 6709
id: b'1622145940100-0', size: 6037
id: b'1622145940266-0', size: 6468
id: b'1622145940435-0', size: 11607
id: b'1622145940597-0', size: 5577
id: b'1622145940768-0', size: 6041
id: b'1622145940926-0', size: 5570
id: b'1622145941098-0', size: 5916

Learn more

• Redis plug-ins for Grafana Documentation
• My Other Stack Is RedisEdge
• RedisEdge Real-time Video Analytics

Contributing

• Fork the repository.
• Find an issue to work on and submit a pull request.
• Could not find an issue? Look for documentation, bugs, typos, and missing features.

License

• Apache License Version 2.0, see LICENSE.