We want to create a model for a binary classification problem with $10$ x $10$ input grayscale $($single$-$channel$)$ images. For this we choose to use a convolutional neural network, with the following architecture:

A convolutional layer with one convolutional filter with filter$/$kernel size of $3$ x $3,$ stride of $1,$ and no padding;

A max$-$pooling layer for down sampling over $4$ x $4$ patches, with stride $1,$ and no padding;

A fully$-$connected layer with layer size $($number of neurons$)3,$ each using a ReLU activation function; and

An output layer with one neuron, having a sigmoid activation to produce an output probability Pr$($Y$=1|$x$)$

No neurons have bias parameters $($for simplicity$),$ and no layers have an padding $($padding is zero throughout$).$

How many parameters $($weights$)$ does this model have?