A user has collected a lot of $100100$ size images. Some of them contain images of mountains and other airplanes. The issue is that the number of images is large and the user keeps adding more and more images. He did not want to manually label images as mountains or airplanes. He has devised a neural network to do the task. All activation functions are RelU except at the final layer, where it is sofmax. Kemel size for the CNN is throughout kept $55.$ Network architecture is given below $($Note: CNN and Max pool layers are shown in pink and yellow respectively. Blue ones are fully connected layers$).$ Determine the number of trainable parameters for this architecture? Assume bias terms also need to be added in a CNN and image size reduces after convolution operation.

$[$S Marks$]$