Backport PR #3936 on branch v0.14.x (Fix float32 support in denoise_bilateral and denoise_tv_bregman)#3937
Conversation
…al and denoise_tv_bregman
|
hmmm, i'm supposed to upload 0.14.3, do I still do that? |
|
Well, at least you should wait for me to fix my Python-3-only syntax. =P In the past, when the tag on GitHub was all that was "released", we've just overwritten the tag. (ie if it didn't hit PyPI it didn't exist.) Up to you how you prefer to play it though. (I did skip a patch release once for similar reasons.) |
|
yeah, mostly i want to avoid git tag conflicts with core devs that might have done |
|
Honestly, it isn't a core dev that i'm worried about, more first time devs that would have cloned the repo in this last week would now be left with very strange git error messages..... |
|
A very fair point @hmaarrfk! Let's just go with 0.14.4 directly then. I suggest patching the release notes so that 0.14.3 never happened though. |
|
I agree with that idea. I felt like this was one of the major bugs we wanted to backport and I think this helps ensure a smooth upgrade procedure for anybody that might be affected by it. |
|
Is censure not thread safe? __________________ test_keypoints_censure_moon_image_octagon ___________________
@test_parallel()
def test_keypoints_censure_moon_image_octagon():
"""Verify the actual Censure keypoints and their corresponding scale with
the expected values for Octagon filter."""
detector = CENSURE(mode='octagon')
# quarter scale image for speed
detector.detect(rescale(img, 0.25,
multichannel=False,
anti_aliasing=False,
mode='constant'))
expected_keypoints = np.array([[ 23, 27],
[ 29, 89],
[ 31, 87],
[106, 59],
[111, 67]])
expected_scales = np.array([3, 2, 5, 2, 4])
> assert_array_equal(expected_keypoints, detector.keypoints)
E AssertionError:
E Arrays are not equal
E
E (shapes (5, 2), (0, 2) mismatch)
E x: array([[ 23, 27],
E [ 29, 89],
E [ 31, 87],...
E y: array([], shape=(0, 2), dtype=int64)
skimage/feature/tests/test_censure.py:81: AssertionErrorcould be scipy??? |
Backport PR #3936: Fix float32 support in denoise_bilateral and denoise_tv_bregman