Shield:
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
This repo contains code to run the AI Baby Sleep Tracking service as well as a web application which provides the user with analysis and charts based on the recorded sleep data.
- Camera which supports RTSP
- Compute accessible via HTTP requests (I used a Raspberry Pi, but you can use any computer)
- Python 3.10 or lower (3.11 is not supported by MediaPipe [see])
There are two components to configure:
- A sleep tracking python script
- A web application
I don't treat these projects as I would in industry. The result is a very monolithic and duct-tapey project.
If there is a desire, I will refactor and make this more repeatable and easy to use.
Install requirements: pip install -r requirements.txt
Run: python main.py
That's it. Except this is where the fun starts.
Most of the dependencies are self explanatory, the only issue I had was installing MediaPipe on Raspbian. I believe I used https://pypi.org/project/mediapipe-rpi4/, but I ran into a number of other issues I won't document here. glhf
There are number of environment variables and holes you'll need to fill with info about your environment. Instead of fixing things, I left a lot of comments.
Alternatively you can touch .env and then copy and paste the contents of .env_sample into it. Then fill in the blanks.
The sleep data is written to sleep_logs.csv. I primed this file with a few rows as an example. Feel free to remove these and start from scratch.
This one is more straight forward. Just make sure you have yarn.
Execute the following commands:
cd webapp; yarn install; yarn start;
And you'll probably get a warning about the app trying to boot on port 80. You can change it to whatever you want in the package.json.
You'll need to update some paths and IPs in the code.
It is also possible to install the app into a docker container. Clone the repository. Navigate to the path and build the docker image.
docker image build -t babysleepcoach .
Run the container (adjust the ports, URLs and path).
With an environment file (for the .env file see env_sample_docker):
docker run -d -p7080:80 -p7081:8000 --env-file .env --name babysleepcoach babysleepcoach
Or with the parameters direct via command line:
docker run -d -p7080:80 -p7081:8000 -e REACT_APP_BACKEND_URL=URL_OF_THE_DOCKER_CONTAINER:7081 -e CAM_STREAM_URL=rtsp://user:password@URL_OF_CAMERA:554/stream1 -e DEBUG=False -e OWL=False -e HATCH_IP=127.0.0.1 -v /PATH_TO_SLEEP_LOGS_ON_HOST/sleep_logs.csv:/usr/app/babysleepcoach/sleep_logs.csv --name babysleepcoach babysleepcoach
The frontend can be accessed via http://URL_OF_THE_DOCKER_CONTAINER:7080 (or the port you wrote in the run command).