1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
|
# See "Writing benchmarks" in the asv docs for more information.
# https://asv.readthedocs.io/en/latest/writing_benchmarks.html
import numpy as np
from skimage import color, data, feature, util
class FeatureSuite:
"""Benchmark for feature routines in scikit-image."""
def setup(self):
# Use a real-world image for more realistic features, but tile it to
# get a larger size for the benchmark.
self.image = np.tile(color.rgb2gray(data.astronaut()), (4, 4))
self.image_ubyte = util.img_as_ubyte(self.image)
self.keypoints = feature.corner_peaks(
self.image, min_distance=5, threshold_rel=0.1
)
def time_canny(self):
feature.canny(self.image)
def time_glcm(self):
pi = np.pi
feature.greycomatrix(
self.image_ubyte, distances=[1, 2], angles=[0, pi / 4, pi / 2, 3 * pi / 4]
)
def time_brief(self):
extractor = feature.BRIEF()
extractor.extract(self.image, self.keypoints)
def time_hessian_matrix_det(self):
feature.hessian_matrix_det(self.image, 4)
|
