Glasses-prediction

About dataset

The dataset has been taken from this website https://www.kaggle.com/jeffheaton/glasses-or-no-glasses?select=faces-spring-2020.

About model architecture

layers : Conv2D--->MaxPool2D---->Conv2D--->MaxPool2D---->Conv2D--->MaxPool2D---->Flatten--->Dense---->Dropout--->Dense
activation : (relu) (relu) (relu) (relu) (sigmoid)
fliter : 32,(3,3)
pool_size = (2,2)
input_shape = (48,48,1)
output_shape = 128

Metrics

loss='binary_crossentropy'
optimizer='adam'
metrics=['accuracy']

Save architecture

I have used h5 format for saving.

Used the trained model glass.h5

from keras.models import load_model
model = load_model('glass.h5')
model.predict_classes(image)