why scipy imresize

Question

why scipy imresize

Closed this issue 3 years ago · 6 comments

Hi,

Any reason to prefer scipy version over OpenCV, skimage or PIL?

Answer 1 · 2021-07-20T17:01:37.000Z

I don't think so. @jiangwei221 why not PIL (which is my favourite). @ducha-aiki Is the version Wei using having any bugs here?

Answer 2 · 2021-07-20T17:07:49.000Z

No idea - I am trying to run it locally on my laptop, where I cannot install required scipy version. So far I am replacing it with skimage.transform.resize and skimage.transform.rescale in order to run it.

Answer 3 · 2021-07-20T17:10:28.000Z

It works with skimage, at least on the handlabeled keypoints from the WxBS ground truth ;) and I managed to run it without cuda!

Answer 4 · 2021-07-20T17:43:41.000Z

Hi, I've removed the dependency on misc.imresize, and replaced with cv2.resize. There is no specific reason that I choose imresize over others. I checked the values for both uint8 resize and float number resize, and the numbers match.

Answer 5 · 2021-07-23T20:41:21.000Z

Nah, cv2.resize != misc.imresize. My previous check must be wrong.

Answer 6 · 2021-07-23T22:27:06.000Z

Moved to PIL