without cnn, if we use fully connected input layer of 1000 neurons for a 100x100 rgb image, how many parameters we will need for that layer?