SomniAI helps improve your sleep quality by analyzing your sleep posture and movements, delivering personalized feedback that helps you wake up refreshed and ready to take on the day. With higher-quality sleep, you can wake up feeling refreshed and energized to start your day. Camera data collected through the mobile app is transmitted to a FastAPI server, where an AI model processes and analyzes the information in real time.

MoJI is an application that captures your sleep posture through a camera and sends the data to a server for movement analysis, helping guide you toward better sleep posture.

For deeper description how we made MoJI app, please read MoJI Documents.

[Mobile App (MoJI)] ---> [AI Server] ---> [Sleep Feedback (HW)]

We made AI Server using FastAPI. Please check out here

In order to use SomniAI, Please follow below instruction to install.

1. Download Dataset (Optional)
In oder to test SomniAI, you can download dataset via kaggle - IEEE VIP CUP 2021 Dataset or below bash api.

kaggle datasets download awsaf49/ieee-vip-cup-2021-train-val-dataset

Or use our Python script:

python data/download_data.py --path /path/to/dataset

2. Run FastAPI Server

./deploy_private.sh

3. Run Nest.js and Next.js Web Server

./deploy_public.sh

You can test the server with sample data using:

python tests/test.py 

You can train ViTPose with IEEE 2021 VIP CUP Dataset using:

./mmpose/run_train.sh

MIT License

[email protected]

Our implemnation is mainly based on the following codebases. We gratefully thank the authors for their wonderfull works.

MMPose
ViTPose
react-native-vision-camera
react-native-fast-opencv
docker-android