Week 2 Classic networks AlexNet 8:40

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

I am probably missing a detail here where
13x13x384 volums is convolved with a 3x3 filter resulting in 13x13x384 filter

then again volume of the same size 13x13x384 is convolved with 3x3 filter. yet we get 13x13 x256 volume.

Is there a detail on stride or padding choice possibly missing in the video?

2

They are evidently using “same” padding and stride of 1. Note that the number of “channels” (the last dimension) is purely a choice: it depends on the number of filters specified in the given layer.