Every 2 seconds take a photo with the webcam, and then send a notification (Mac) if nail biting is detected.
Powered by a cnn, currently trained on a the dataset linked below. This dataset is diverse, but biased towards myself, the model kinda generalizes but adding yourself to the dataset would make it much more effective.
pip install git+https://github.com/alecsharpie/nomo_nailbiting
monitor_nailbiting
#OR
python -m nomo_nailbiting.main
This process is shown in the Create_Dataset.ipynb
notebook.
Available here: https://huggingface.co/datasets/alecsharpie/nailbiting_classification
Used a binary image dataset for classifying nailbiting. Images are cropped to only show the mouth area. Should contain edge cases such as drinking water, talking on the phone, scratching chin etc.. all in "no biting" category.
The data was scraped from stock images sites and photos of myself were taken with my webcam. MTCNN (https://github.com/ipazc/mtcnn) was then used to crop the images down to only the show the mouth area The images were then converted to a black & white colour scheme.
Check out Train_Model_WandB.ipynb
.