Question $2$

Assume that you have RGB images of spatial dimension $28$x$28.$ The images are processed by a network consisting of a backbone and a classifier used for binary classification. The backbone consists of $2$ blocks. Specifically, the model architecture is described below:

$1.$ First block. The first block comprises three consecutive layers: a$)$ a convolutional layer with $163$x$3$ kernels and a stride of $1,$ and padding of $1$ b$)$ a max$-$pooling layer with a $2$x$2$ kernel and a stride of $2($which halves the spatial dimensions of the input tensor$),$ and c$)$ a ReLU activation function. The sequence of operations is: convolution, ReLU activation, max$-$pooling.

$2.$ Second block. The second block contains a convolutional layer with $323$x$3$ kernels, a stride of $1,$ and padding of $1,$ followed by a ReLU activation and a max$-$pooling layer with a $2$x$2$ kernel and a stride of $2.$

$3.$ Classifier. The output of the second block is flattened. The flattened tensor is passed through a Multi$-$Layer Perceptron $($MLP$)$ classifier with three fully connected layers. ReLU activation functions are applied after the first two fully connected layers. The first linear layer maps the input features to $128$ features. The second linear layer maps these features to $64$ features. The third linear layer maps the $64$ features to a single value.

Assume that all layers do not have bias terms. What is the total number of parameters in this network? Provide all the details of your calculation.