Inspiration
We were talking about the reason for the rise in coronavirus cases here in the US. Then we thought about how convenient it would be to have a security system that automatically warns a person if they don't have a mask on.
What it does
It detects whether the person in the picture has a mask, mask on but wore incorrectly, or no mask at all.
How I built it
I trained the model using a dataset I found on Kaggle. After tuning the hyperparameter a little bit and trying out different models, I was able to get a satisfying result.
Challenges I ran into
The batches for training were too large, thus my GPU could not handle the training. The training took overall 10 hours, and the result was somewhat inaccurate so I had to retrain it using a different model with an accuracy matrix in mind.
Accomplishments that I'm proud of
After 18-20 hours of training, I was finally able to see a satisfying accuracy from my trained model.
What I learned
Instead of testing the model after hours of training, I should select the type of algorithm that better suits my data beforehand.
What's next for Flask Mask
Instead of being on a website, this model can be implemented in a security system, and provide safer environments in populated areas such as airports and markets.
Log in or sign up for Devpost to join the conversation.