Model

Data preparation

1. Download MaskedFace-Net dataset from https://github.com/cabani/MaskedFace-Net
2. Move the content to get the following directory structure:

├── ai-mask-model
│   ├── data                        <- All your data modules should be located here!
│   │   ├── dataframe               <- Place to store JSON with mapping data calsses
│   │   │
│   │   └── dataset                 <- Whole dataset
│   │       ├── test                
│   │       │   ├── 0               <- Part of the dataset containing unproperly masked faces
│   │       │   │   ├── 00001.jpg   <- Image
│   │       │   │   ├── 00002.jpg
│   │       │   │   └── ...
│   │       │   └── 1               <- Part of the dataset containing properly masked faces
│   │       │       ├── 00001.jpg   <- Image
│   │       │       ├── 00002.jpg
│   │       │       └── ...
│   │       ├── train
│   │       │   ├── 0               <- Part of the dataset containing unproperly masked faces
│   │       │   │   ├── 00001.jpg   <- Image
│   │       │   │   ├── 00002.jpg
│   │       │   │   └── ...
│   │       │   └── 1               <- Part of the dataset containing properly masked faces
│   │       │       ├── 00001.jpg   <- Image
│   │       │       ├── 00002.jpg
│   │       │       └── ...
│   │       └── valid
│   │           ├── 0               <- Part of the dataset containing unproperly masked faces
│   │           │   ├── 00001.jpg   <- Image
│   │           │   ├── 00002.jpg
│   │           │   └── ...
│   │           └── 1               <- Part of the dataset containing properly masked faces
│   │               ├── 00001.jpg   <- Image
│   │               ├── 00002.jpg
│   │               └── ...
│   └── ...
└── ...

How to run

First, install dependencies

# clone project
git clone https://github.com/HTD-Health/ai-smart-mirror.git

# install project
cd ai-smart-mirror
pip install -r requirements.txt

Next, just train the model by running:

python train.py data/datasets/

When you finish training your network, to use it just type:

python predict.py path_to_image_to_check.jpg

Raspberry Pi 3

Remote access

We are using Dataplicity for remote access to Raspberry Pi. If you want to have access to Raspberry Pi for test purposes send your email address and request to be added to Jędrzej Polaczek.

How to use

Repository

Repository is under path home/pi/Workplace/ai-smart-mirror

Making changes

To make changes (like changing branch) you need to run the command as superuser. To make it be you need to type su pi (and execute this command) and after that enter the appropriate password. Quick guide: https://docs.dataplicity.com/docs/superuser

How to install

To install this library just run in ai-mask-model folder command pip install -e . when you are in the right virtual environment.