MaskDetection-COVID-19
Introduction
MaskDetection-COVID-19 is a project that aims to solve the problem of identifying if someone has worn a mask or not. It is based on Convolutional Neural Network Architecture(CNN) and it is trained and integrated with computer vision technology to actively identify that if some user has worn a mask or not.
Neural Network Architectural Details:
- First Hidden Layer:
- Convolution Layer: 200 Neurons(3X3)
- Activation Function: RELU
- Max Pooling Layer (2X2)
- Second Hidden Layer:
- Convolution Layer: 100 Neurons(3X3)
- Activation Function: RELU
- Max Pooling Layer (2X2)
- Final Layer:
- Dense Layer: 50 neurons
- Dropout
- Output Layer:
- Dense: 2 neurons
- Activation: Softmax Transform
- Loss: Categorical Crossentropy
- Optimizer: Adam
- Model Results:
- Accuracy: 0.9492
- Loss: 0.1332
Usage:
Clone the repo and run the .ipynb notebooks in the order of
- dataPreprocessing.ipynb
- CNNTraining.ipynb
- maskDetection.ipynb
The code is well-commented and the details regarding implementation, CNN architecture related details are deliberated. For further reference, check out the snippets of code from Image Recognition, and CNN with Keras and Tensorflow repositories.
Libraries used:
- cv2 (Computer Vision Library)
- Tensorflow + Keras
- Numpy
- sklearn
- Matplotlib