CNN layers interpretation - Course 4 - Week 1- CNN example

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

So in the lecture, we said that a Conv_1+ Pooling_1 is considered as Layer 1 instead of layer 1 and layer 2. What about flatten() and Dense layers?

image
image2064×547 63 KB

Is my interpretation correct?

2

Why does it matter what you call them? If it were up to me, I don’t think the “flatten” layer is worth giving a number to: nothing really happens there and there are no parameters affected. But each of the Dense layers is clearly a layer on its own. Maybe you bundle the “flatten” with either the first Dense layer or the previous Conv + Pool layer.

3

Note that this is just a naming convention: it has nothing to do with how forward propagation or back propagation actually work, right? That’s why I asked “why does it matter?”

4

Oh yeah, I was asking just for the naming convention. Just to be on the same page with everyone else. In the end, it doesn’t really matter what you call them.