This question evaluates your knowledge about CNN architectures and classification.

For all convulation layers, the Filter size must be (nxn) for example=n=5, then filter size=5x5

Paddding=0, stride=1 you can choose anohter padding and stride ratios.

a(20 ) Assume that you have 5 types of leaf diseases. Each image (grayscale) is in the form 224x224 pixels. Design a LeNet5 like architecture for training this dataset and generating predictions. Your LeNET architecture must include, Convolution (4), Activation(relu), MaxPool, Dropout, Softmax, Fully Connected Layer1 (4096), Fully Connected Layer2 (2048). You must use at least 4 convolution layers. The number of other layers depend on your selection.

b(20 ) Assume that you have 40 types of leaf diseases. Each image (color, RGB) is in the form 224x224x3 pixels. Design a LeNet5 like architecture for training this dataset and generating predictions.Your LeNET architecture must include, Convolution (4), Activation(relu), Dropout, Softmax, Fully Connected Layer1 (2048), Fully Connected Layer2 (256). You must use at least 4 convolution layers. The number of other layers depend on your selection.