1
$\begingroup$

I'm currently reading this explanation of convolutional neural networks and there's a part around strides that I don't quite understand. I'm just starting with this, so I apologize if this is a really basic question. But I'm trying to develop an understanding and some of these images have thrown me off.

Specifically, in this image

enter image description here

The stride has been increased to 2 and it's using a 3x3 filter (represented by the red, green and blue outline squares in the first picture)

Why is the blue square below the red one and not shifted to the side of the green one ending at the edge of the 7x7 volume? Should it not move left to right then down 2 squares when it reaches the next line?

I'm not sure if the author is just trying to show the stride moving down as it goes, but I think my confusion stems from the fact that the 1 stride image example is only moving in the horizontal direction (as seen below).

enter image description here

Is there something fundamental I haven't grasped here?

$\endgroup$
1
$\begingroup$

Yes, you're right, after the green one, it should also move two steps (because stride = 2) to the right once more. Note that in the $3 \times 3$ output volume picture, there's also still a white cell in the top right corner. That cell would get filled with whatever colour you choose to move to the right after the green one.

The blue one would then follow after what I described above, as the fourth square. I guess the author simply didn't feel like drawing a third square to the right of the green one, because the red + green squares already illustrate how the pattern works. The blue one was probably additionally drawn to illustrate that stride works the same way vertically as it does horizontally, e.g. blue also moves down two rows (because stride = 2).

See the bottom of the picture in the second answer here.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.