cd30e010d5
Also added a nogil context manager to hoghistograms
2015-06-19 16:01:52 +02:00
691beaa59f
Adding nogil to cell_hog
2015-06-19 15:49:48 +02:00
51e683f665
I hadn't realized I was reusing cc in the loop, so it would crash after one iteration.
...
And I had changed the second orientation, which probably wasn't correct
2015-06-19 08:51:54 +02:00
167c8eeeb6
I hadn't realized when I changed orientations to an array, that its actually just an int and the original code said range(orientations), so I changed it to np.arange(orientations). Hopefully the tests will now pass again
2015-06-18 22:08:34 +02:00
2500c21b12
Converted orientation and magnitude to typed memory view. Also changed gradient_columns and gradient_rows, hopefully in the right way.
2015-06-18 22:07:28 +02:00
3d85503472
Renamed x to rows and y to columns, though I might have still missed it in some spots.
...
Indented the long if statements and put continue on a new line
Precomputed the range start and stop values
Defined additional variables
2015-06-18 22:03:21 +02:00
bea9886a12
I noticed that cy+cy and cx+cx was repeated in the inner-most loop as well
2015-06-18 22:01:44 +02:00
ded652c2ab
Moved more computations out of loops and added division from future
2015-06-18 18:55:32 +02:00
f633262f03
Moved more variables out of the loops and modified the docstring
2015-06-18 18:50:34 +02:00
9c141eadf8
Extracted row_index and cell_row from the inner loop
...
Defined constants that don't have to be recomputed several times
2015-06-18 18:45:29 +02:00
1b9a15a4d4
Made the indents line up with the (
2015-06-18 18:38:59 +02:00
03b7d638a7
Wrapped long lines according to pep8.
2015-06-16 09:57:49 +02:00
605eac2e8c
Renamed to use row/column naming convention.
2015-06-16 09:54:54 +02:00
78cf5a0f9a
Removed unused code.
2015-06-16 09:40:04 +02:00
8ab6e63e75
Removed unused import, applied pep8 formatting and corrected a method call
...
that should have been refactored.
2015-06-16 09:36:02 +02:00
c37fc9ec20
Update _hog.py
...
Made hog_histograms snake case
2015-06-15 22:51:30 +02:00
502377e9ce
Update _hoghistogram.pyx
...
Made the function names PEP8 compatible and added another level of indentation.
I also tried renaming some things where I considered columns useful, but for example cells per row isn't entirely accurate. Any suggestions?
2015-06-15 22:49:20 +02:00
ecdf71d14d
Switched to memory view syntax.
2015-06-15 15:41:18 +02:00
aedf63f1f7
Switched to cnp where applicable.
2015-06-15 15:25:30 +02:00
3719ff216c
Updated docstrings.
2015-06-15 13:14:34 +02:00
c756d2d492
Partially fulfilled code review.
2015-06-15 11:49:21 +02:00
2e56e685d8
change import syntax for python 3
2015-06-15 11:05:07 +02:00
d55135ddbb
Fix comments and tab whitespace
2015-06-15 11:03:58 +02:00
f09907961a
Initial attempt at optimising HOG features with cython
2015-06-15 11:03:57 +02:00
0d134987f9
Harmonize all ndimage usage across the library
...
Only two forms remain in use:
- `from scipy import ndimage as ndi`
- `from scipy.ndimage import function`
2015-06-09 15:18:37 +10:00
990892fcde
Replace libc.math.NAN with numpy.math.NAN for MSVC
2015-05-21 08:27:00 -07:00
a2ddba21a5
Do not acquire GIL for hessian_matrix_det
2015-05-20 20:22:02 -07:00
4a95345489
Do not acquire GIL for texture functions
2015-05-20 17:51:41 -07:00
9cd7ca5881
Do not acquire GIL for BRIEF
2015-05-20 17:39:46 -07:00
bca458bc1c
Do not acquire GIL for CENSURE
2015-05-20 09:04:58 -07:00
a489b85b22
Do not acquire GIL for corner detectors
2015-05-20 07:45:04 -07:00
c52c7f4bfb
Do not acquire GIL for ORB
2015-05-20 07:38:47 -07:00
d5099db362
Fix PEP8
2015-05-03 01:32:56 +02:00
b80e494484
Change data types to conform to style guide
2015-05-02 18:12:30 +02:00
dc0a91b569
Fix doctests for #1257
...
Changing the return type of the blob functions from integer to float
arrays invalidated the docstring examples. Updating the examples fixes
the doctests.
2015-05-02 18:12:30 +02:00
c86c6effe8
Fix tests for #1257
...
The test of blob_doh() set an unrealistically strict threshold for the
accuracy of the estimated blob radius, causing the test to fail with the
true floating-point value (previously, the test passed only because
sigma was truncated to integer). Increasing the threshold fixes the
problem.
2015-05-02 18:12:30 +02:00
833edc17ef
Fixed typo
2015-05-02 18:12:30 +02:00
eb6ddc735b
Changed local_maxima to float32
2015-05-02 18:12:30 +02:00
43246e30af
Blob detection supports for non-integer sigmas
2015-05-02 18:12:30 +02:00
d6870c6de7
Merge pull request #1066 from msarahan/PhaseCorrelation
...
Add subpixel shift registration
2015-03-02 08:30:45 -08:00
84bcb583d7
Fix Hessian matrix eigvals
2015-02-23 22:32:04 -05:00
71a89697f0
improve test coverage. Use SciPy's fourier_shift for examples.
2015-02-21 11:33:37 -08:00
cf09b0c5ef
Incorporate fixes for complex images
2015-02-20 20:55:46 -08:00
ed7aecdc4c
Fix links
2015-02-07 19:33:30 -06:00
be7154f23d
Skip build on 3.2
...
Skip doc build on 3.2
Skip doc build on 3.2
Use block literal
Fix html make target
Revert change to threshold_isodata
Fix skimage.novice and thresholding warnings
Fix feature.__init__ __all__
Fix generic.py link
Doc reference fixes
Fix missing class member warnings
Add parallelization to the user guide toc
Remove unused automodule directives
Fix sub_dirs in gallery_index
Remove unused file and fix reference to api/api
Add missing links to user guide
Use pre-output links to images
Recover from Juan's doc 'upgrades'
2015-02-07 16:40:37 -06:00
03dc489457
Add subpixel-precision image translation registration function to feature module
2015-01-29 20:09:50 -08:00
b6a1bcdb06
Merge pull request #1324 from ahojnnes/rel-import
...
Use relative imports in skimage files
2015-01-25 15:26:34 -06:00
305f5fd225
Using numpy.argwhere makes code easier to read
2015-01-25 00:22:11 +00:00
c5ac705123
Fixed bug in feature.peak_local_max: num_peaks should now be usable for ND arrays.
2015-01-25 00:19:12 +00:00
22ffc20fc4
Added broken test for feature.peak_local_max when the image is 3D and num_peaks is set. The current code only holds for 2D images.
2015-01-25 00:15:20 +00:00
Ivo Flipse