Categroized images- 4 types of subtypes of Blood cells CATEGORIES=["EOSINOPHIL","LYMPHOCYTE","MONOCYTE","NEUTROPHIL"] train data - 9957 img test data - 2487 img
CNN architecture --- 3 convolutional layers - each with 3x3 kernel size and filter values 32, 64 and 128. Each layer is seperated by an activation layer(Relu).]
Followed by 3 Dense layers with activation function Relu. Final Dense is activated with softmax for 4 category prediction.