The dataset used for this project is from github user X-zhangyang's Real-World-Masked-Face-Dataset repository.
Precisely, this one -> Click for the dataset <-
The dataset offers 90,000 face images without masks, 2203 face images with masks. In order to balance the mask/no_mask ratio and eliminate memory issues, we randomly removed 87,800 face images without masks. Then we trained the model on the remaining 4383 images for mask - no_mask classification.
train_mask_detection_model.ipynb: The notebook we used to train our model.
detect_mask_on_image.ipynb: The notebook we tested our model on images.
detect_mask_on_video.ipynb: The notebook we tested our model on the live video captured by our webcams.
remove_images_from_dataset.py: We had issues with memory on the training the full Real-World-Masked-Face-Dataset, hence, we used this method to randomly remove image files from the dataset.
self-built-masked-face-recognition-dataset.zip: The dataset we used for the training the model. It's derived from the Real-World-Masked-Face-Dataset, with the usage of remove_images_from_dataset.py.
CV_Project_Report.pdf: Our report can be found in the folder "project_report".
Note: All the sources we used building this project, weather it's for the code or the report, can be found on the "References" part of our report.