MobileNet v2 Number of Parameters

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

Hi I understand the architecture of MobileNet v2 (input → expansion → depth wise → point wise → …) but how do we compute the total number of parameters for this model? Assuming that the input is of size n x n x 5, expansion has 30 filters, followed by 3 x 3 depth wise filter, and 20 filters in projection.

My only confusion here is the computation of the total number of parameters when n is unknown. Could someone help me out?

Thanks!

2

The point is that the size of the input in a ConvNet in terms of the h and w dimensions has no impact on the number of parameters, right? The parameters are the elements of the filters and the size of the filters only depend on the number of channels in the input, not the number of h and w pixels.

1 Like