Every Position, at Every Size

The first idea anyone has for detection is the obvious one: slide a box over the image and classify every crop it lands on.

You are doing exactly that on one $256 \times 256$ image.

setting	value
window sizes	$32 \times 32$ , $64 \times 64$ , $128 \times 128$
stride	$8$ pixels, across and down
first stop	flush against the top-left corner
rule	a window may never hang over an edge — every stop must lie entirely inside the image

Every stop of every window, at every size, is one run of the classifier.

A two-stage detector works on the same image differently: it proposes 300 candidate regions and classifies only those.

How many times as many classifier runs does the sliding window need as the two-stage detector?

Round your answer to 2 decimal places.