RedisGears / AnimalRecognitionDemo

An example of using Redis Streams, RedisGears and RedisAI for Realtime Video Analytics (i.e. filtering cats)

AnimalRecognitionDemo

This demo combines several Redis data structures and Redis Modules to process a AnimalRecognitionDemostream of images and filter out the images that contain cats.

It uses:

• Redis Streams to capture the input video stream: all
• RedisGears to process this stream
• RedisAI to classify the images with MobilenetV2

It forwards the images that contain cats to a stream: cats

It uses RedisAI Integration in RedisGears with asynchronous function so the server is not blocked while RedisGears is triggering an inference session in RedisAI.

Architecture

Requirements

Docker and Python 3

Running the Demo

To run the demo:

git clone https://github.com/RedisGears/AnimalRecognitionDemo.git
cd AnimalRecognitionDemo
# If you don't have it already, install https://git-lfs.github.com/ (On OSX: brew install git-lfs)
git lfs install && git lfs fetch && git lfs checkout

For running the demo with make, run:

make start
make camera

Then open the UI to watch the result streams.

To end the demo, then to stop the containers:

make stop

Run make help for a few more options.

For running the demo manually, run:

docker-compose up

If something went wrong, e.g. you skipped installing git-lfs, you need to force docker-compose to rebuild the containers

docker-compose up --force-recreate --build

Open a second terminal for the video capturing:

pip install -r camera/requirements.txt
python camera/read_camera.py

Or run the camera process in test mode (without streaming from your camera):

ANIMAL=[cat|dog] python camera/read_camera.py --test

UI

• http://localhost:3000 shows all the captured frames
• http://localhost:3001 shows only the framse with cats

Limitations

This demo is designed to be easy to setup, so it relies heavily on docker. You can get better performance and a higher FPS by runninng this demo outside docker. To control the FPS, edit the gear.py file.