Chest XRAY Images(Source - Kaggle)
This dataset contains Chest XRAY images with 2 classes - healthy Chest XRAY Images , Pneumonia Chest XRAY Images
Sample Image :
The model is based on Encoder-Decoder networks .
The Encoder downsamples the image with two key differen hidden layers : mu(Learned mean) and sigma(standard deviation)
The decoder Network upsamples the samples reparameterised from the outputs of the encoder(mu and sigma)
Optimizer used : Adam
Loss functions : KL Divergence Loss and Binary Cross Entropy Loss
The model was trained for 50 Epochs with learning rate of 0.003
And if the loss is constant or increasing for 3 iterations then the model is assumed to be trained!