Through the use of convolutional neural networks, I made a model that it's capable of detecting by reading an image if a person is wearing a face mask or not. The database I'm using contains about 6400 images of people. Each of these images also has a JSON type file that indicates many of its characteristics, being "class_name" the one that indicates the types of accessories that the people in the image are wearing (where we're going to classify if they're wearing a mask or not). The objective of this project is to apply CNN for a possible tool that uses computer vision and artificial intelligence to prevent infections in closed spaces.
I created a notebook where I do the data preprocessing and model building. It is important to identify the characteristics of the dataset and the transformations that I applied:
- Class name extract by JSON file
- Data separation and normalization (124x124 size)
- Data augmentation with ImageDataGenerator from keras image preprocessing
- Use of scikit-learn for data treatment
The model architecture is the following:
The model has 23M training parameters aprox and you can download it here
Made by @yepedraza