wehub-resource-sync
268 lines
14 KiB
Python
268 lines
14 KiB
Python
"""
|
|
Tests to measure the performance of each augmenter.
|
|
Run these checks from the project directory (i.e. parent directory) via
|
|
python check_performance.py
|
|
"""
|
|
from __future__ import print_function, division
|
|
import imgaug as ia
|
|
from imgaug import augmenters as iaa
|
|
import numpy as np
|
|
import time
|
|
import random
|
|
import six.moves as sm
|
|
|
|
"""
|
|
---------------------------
|
|
Keypoints
|
|
---------------------------
|
|
[Augmenter: Identity]
|
|
(4, 4, 3) | SUM 0.01990s | PER ITER avg 0.00020s, min 0.00017s, max 0.00043s
|
|
(32, 32, 3) | SUM 0.01863s | PER ITER avg 0.00019s, min 0.00017s, max 0.00033s
|
|
(256, 256, 3) | SUM 0.01879s | PER ITER avg 0.00019s, min 0.00017s, max 0.00029s
|
|
[Augmenter: Crop-px]
|
|
(4, 4, 3) | SUM 0.20215s | PER ITER avg 0.00202s, min 0.00168s, max 0.01908s
|
|
(32, 32, 3) | SUM 0.19844s | PER ITER avg 0.00198s, min 0.00164s, max 0.01933s
|
|
(256, 256, 3) | SUM 0.17918s | PER ITER avg 0.00179s, min 0.00166s, max 0.00214s
|
|
[Augmenter: Crop-percent]
|
|
(4, 4, 3) | SUM 0.14201s | PER ITER avg 0.00142s, min 0.00114s, max 0.02041s
|
|
(32, 32, 3) | SUM 0.16912s | PER ITER avg 0.00169s, min 0.00137s, max 0.02023s
|
|
(256, 256, 3) | SUM 0.15548s | PER ITER avg 0.00155s, min 0.00142s, max 0.00193s
|
|
[Augmenter: Fliplr]
|
|
(4, 4, 3) | SUM 0.02303s | PER ITER avg 0.00023s, min 0.00021s, max 0.00034s
|
|
(32, 32, 3) | SUM 0.02477s | PER ITER avg 0.00025s, min 0.00021s, max 0.00038s
|
|
(256, 256, 3) | SUM 0.02383s | PER ITER avg 0.00024s, min 0.00022s, max 0.00036s
|
|
[Augmenter: Flipud]
|
|
(4, 4, 3) | SUM 0.02362s | PER ITER avg 0.00024s, min 0.00021s, max 0.00035s
|
|
(32, 32, 3) | SUM 0.02356s | PER ITER avg 0.00024s, min 0.00021s, max 0.00032s
|
|
(256, 256, 3) | SUM 0.02415s | PER ITER avg 0.00024s, min 0.00021s, max 0.00037s
|
|
[Augmenter: Grayscale]
|
|
(4, 4, 3) | SUM 0.01908s | PER ITER avg 0.00019s, min 0.00017s, max 0.00030s
|
|
(32, 32, 3) | SUM 0.01903s | PER ITER avg 0.00019s, min 0.00017s, max 0.00030s
|
|
(256, 256, 3) | SUM 0.01876s | PER ITER avg 0.00019s, min 0.00017s, max 0.00027s
|
|
[Augmenter: GaussianBlur]
|
|
(4, 4, 3) | SUM 0.01904s | PER ITER avg 0.00019s, min 0.00017s, max 0.00029s
|
|
(32, 32, 3) | SUM 0.01851s | PER ITER avg 0.00019s, min 0.00017s, max 0.00033s
|
|
(256, 256, 3) | SUM 0.01894s | PER ITER avg 0.00019s, min 0.00017s, max 0.00025s
|
|
[Augmenter: AdditiveGaussianNoise]
|
|
(4, 4, 3) | SUM 0.01902s | PER ITER avg 0.00019s, min 0.00017s, max 0.00029s
|
|
(32, 32, 3) | SUM 0.01905s | PER ITER avg 0.00019s, min 0.00017s, max 0.00028s
|
|
(256, 256, 3) | SUM 0.01971s | PER ITER avg 0.00020s, min 0.00017s, max 0.00046s
|
|
[Augmenter: Dropout]
|
|
(4, 4, 3) | SUM 0.01887s | PER ITER avg 0.00019s, min 0.00017s, max 0.00027s
|
|
(32, 32, 3) | SUM 0.01913s | PER ITER avg 0.00019s, min 0.00017s, max 0.00030s
|
|
(256, 256, 3) | SUM 0.01922s | PER ITER avg 0.00019s, min 0.00017s, max 0.00029s
|
|
[Augmenter: Multiply]
|
|
(4, 4, 3) | SUM 0.01942s | PER ITER avg 0.00019s, min 0.00017s, max 0.00028s
|
|
(32, 32, 3) | SUM 0.01922s | PER ITER avg 0.00019s, min 0.00017s, max 0.00032s
|
|
(256, 256, 3) | SUM 0.01875s | PER ITER avg 0.00019s, min 0.00017s, max 0.00030s
|
|
[Augmenter: ContrastNormalization]
|
|
(4, 4, 3) | SUM 0.01852s | PER ITER avg 0.00019s, min 0.00017s, max 0.00029s
|
|
(32, 32, 3) | SUM 0.01869s | PER ITER avg 0.00019s, min 0.00017s, max 0.00026s
|
|
(256, 256, 3) | SUM 0.01875s | PER ITER avg 0.00019s, min 0.00017s, max 0.00028s
|
|
[Augmenter: Grayscale]
|
|
(4, 4, 3) | SUM 0.01919s | PER ITER avg 0.00019s, min 0.00017s, max 0.00030s
|
|
(32, 32, 3) | SUM 0.01923s | PER ITER avg 0.00019s, min 0.00017s, max 0.00033s
|
|
(256, 256, 3) | SUM 0.01888s | PER ITER avg 0.00019s, min 0.00017s, max 0.00028s
|
|
[Augmenter: ElasticTransformation]
|
|
(4, 4, 3) | SUM 0.01882s | PER ITER avg 0.00019s, min 0.00017s, max 0.00024s
|
|
(32, 32, 3) | SUM 0.01883s | PER ITER avg 0.00019s, min 0.00017s, max 0.00029s
|
|
(256, 256, 3) | SUM 0.01869s | PER ITER avg 0.00019s, min 0.00017s, max 0.00029s
|
|
[Augmenter: AffineOrder0ModeConstant]
|
|
(4, 4, 3) | SUM 0.28146s | PER ITER avg 0.00281s, min 0.00243s, max 0.02199s
|
|
(32, 32, 3) | SUM 0.28047s | PER ITER avg 0.00280s, min 0.00243s, max 0.02083s
|
|
(256, 256, 3) | SUM 0.28715s | PER ITER avg 0.00287s, min 0.00243s, max 0.02088s
|
|
[Augmenter: AffineOrder0]
|
|
(4, 4, 3) | SUM 0.27242s | PER ITER avg 0.00272s, min 0.00246s, max 0.00362s
|
|
(32, 32, 3) | SUM 0.29675s | PER ITER avg 0.00297s, min 0.00247s, max 0.02220s
|
|
(256, 256, 3) | SUM 0.28988s | PER ITER avg 0.00290s, min 0.00247s, max 0.02128s
|
|
[Augmenter: AffineOrder1]
|
|
(4, 4, 3) | SUM 0.26750s | PER ITER avg 0.00267s, min 0.00246s, max 0.00321s
|
|
(32, 32, 3) | SUM 0.28361s | PER ITER avg 0.00284s, min 0.00245s, max 0.02144s
|
|
(256, 256, 3) | SUM 0.28973s | PER ITER avg 0.00290s, min 0.00246s, max 0.02070s
|
|
[Augmenter: AffineAll]
|
|
(4, 4, 3) | SUM 0.27070s | PER ITER avg 0.00271s, min 0.00246s, max 0.00367s
|
|
(32, 32, 3) | SUM 0.28405s | PER ITER avg 0.00284s, min 0.00247s, max 0.02120s
|
|
(256, 256, 3) | SUM 0.28895s | PER ITER avg 0.00289s, min 0.00247s, max 0.02144s
|
|
---------------------------
|
|
Images
|
|
---------------------------
|
|
[Augmenter: Identity]
|
|
(16, 4, 4, 3) | SUM 0.00135s | PER ITER avg 0.00001s, min 0.00001s, max 0.00008s
|
|
(16, 32, 32, 3) | SUM 0.00203s | PER ITER avg 0.00002s, min 0.00002s, max 0.00005s
|
|
(16, 256, 256, 3) | SUM 0.05284s | PER ITER avg 0.00053s, min 0.00044s, max 0.00194s
|
|
[Augmenter: Crop-px]
|
|
(16, 4, 4, 3) | SUM 0.09324s | PER ITER avg 0.00093s, min 0.00084s, max 0.00315s
|
|
(16, 32, 32, 3) | SUM 0.10302s | PER ITER avg 0.00103s, min 0.00094s, max 0.00162s
|
|
(16, 256, 256, 3) | SUM 0.81943s | PER ITER avg 0.00819s, min 0.00767s, max 0.00934s
|
|
[Augmenter: Crop-percent]
|
|
(16, 4, 4, 3) | SUM 0.06562s | PER ITER avg 0.00066s, min 0.00057s, max 0.00099s
|
|
(16, 32, 32, 3) | SUM 0.09784s | PER ITER avg 0.00098s, min 0.00089s, max 0.00131s
|
|
(16, 256, 256, 3) | SUM 0.80779s | PER ITER avg 0.00808s, min 0.00732s, max 0.01008s
|
|
[Augmenter: Fliplr]
|
|
(16, 4, 4, 3) | SUM 0.00525s | PER ITER avg 0.00005s, min 0.00004s, max 0.00017s
|
|
(16, 32, 32, 3) | SUM 0.01025s | PER ITER avg 0.00010s, min 0.00007s, max 0.00015s
|
|
(16, 256, 256, 3) | SUM 0.36918s | PER ITER avg 0.00369s, min 0.00181s, max 0.00553s
|
|
[Augmenter: Flipud]
|
|
(16, 4, 4, 3) | SUM 0.00512s | PER ITER avg 0.00005s, min 0.00004s, max 0.00009s
|
|
(16, 32, 32, 3) | SUM 0.00665s | PER ITER avg 0.00007s, min 0.00006s, max 0.00011s
|
|
(16, 256, 256, 3) | SUM 0.12664s | PER ITER avg 0.00127s, min 0.00092s, max 0.00167s
|
|
[Augmenter: Grayscale]
|
|
(16, 4, 4, 3) | SUM 0.05943s | PER ITER avg 0.00059s, min 0.00050s, max 0.00125s
|
|
(16, 32, 32, 3) | SUM 0.12247s | PER ITER avg 0.00122s, min 0.00106s, max 0.00205s
|
|
(16, 256, 256, 3) | SUM 3.62785s | PER ITER avg 0.03628s, min 0.03508s, max 0.03963s
|
|
[Augmenter: GaussianBlur]
|
|
(16, 4, 4, 3) | SUM 0.15514s | PER ITER avg 0.00155s, min 0.00136s, max 0.00188s
|
|
(16, 32, 32, 3) | SUM 0.25121s | PER ITER avg 0.00251s, min 0.00221s, max 0.00298s
|
|
(16, 256, 256, 3) | SUM 5.51685s | PER ITER avg 0.05517s, min 0.04923s, max 0.06026s
|
|
[Augmenter: AdditiveGaussianNoise]
|
|
(16, 4, 4, 3) | SUM 0.09606s | PER ITER avg 0.00096s, min 0.00085s, max 0.00150s
|
|
(16, 32, 32, 3) | SUM 0.21302s | PER ITER avg 0.00213s, min 0.00196s, max 0.00254s
|
|
(16, 256, 256, 3) | SUM 7.22374s | PER ITER avg 0.07224s, min 0.07017s, max 0.07558s
|
|
[Augmenter: Dropout]
|
|
(16, 4, 4, 3) | SUM 0.09362s | PER ITER avg 0.00094s, min 0.00084s, max 0.00118s
|
|
(16, 32, 32, 3) | SUM 0.17472s | PER ITER avg 0.00175s, min 0.00161s, max 0.00230s
|
|
(16, 256, 256, 3) | SUM 5.04969s | PER ITER avg 0.05050s, min 0.04839s, max 0.05631s
|
|
[Augmenter: Multiply]
|
|
(16, 4, 4, 3) | SUM 0.05442s | PER ITER avg 0.00054s, min 0.00046s, max 0.00089s
|
|
(16, 32, 32, 3) | SUM 0.06895s | PER ITER avg 0.00069s, min 0.00060s, max 0.00109s
|
|
(16, 256, 256, 3) | SUM 0.87311s | PER ITER avg 0.00873s, min 0.00799s, max 0.00993s
|
|
[Augmenter: ContrastNormalization]
|
|
(16, 4, 4, 3) | SUM 0.05746s | PER ITER avg 0.00057s, min 0.00050s, max 0.00094s
|
|
(16, 32, 32, 3) | SUM 0.08083s | PER ITER avg 0.00081s, min 0.00071s, max 0.00133s
|
|
(16, 256, 256, 3) | SUM 1.57577s | PER ITER avg 0.01576s, min 0.01443s, max 0.01831s
|
|
[Augmenter: Grayscale]
|
|
(16, 4, 4, 3) | SUM 0.05464s | PER ITER avg 0.00055s, min 0.00049s, max 0.00069s
|
|
(16, 32, 32, 3) | SUM 0.12058s | PER ITER avg 0.00121s, min 0.00104s, max 0.00223s
|
|
(16, 256, 256, 3) | SUM 3.57037s | PER ITER avg 0.03570s, min 0.03461s, max 0.03780s
|
|
[Augmenter: ElasticTransformation]
|
|
(16, 4, 4, 3) | SUM 0.29551s | PER ITER avg 0.00296s, min 0.00272s, max 0.00336s
|
|
(16, 32, 32, 3) | SUM 0.68591s | PER ITER avg 0.00686s, min 0.00642s, max 0.00764s
|
|
(16, 256, 256, 3) | SUM 26.30515s | PER ITER avg 0.26305s, min 0.25754s, max 0.26912s
|
|
[Augmenter: AffineOrder0ModeConstant]
|
|
(16, 4, 4, 3) | SUM 0.35887s | PER ITER avg 0.00359s, min 0.00333s, max 0.00424s
|
|
(16, 32, 32, 3) | SUM 0.47889s | PER ITER avg 0.00479s, min 0.00451s, max 0.00535s
|
|
(16, 256, 256, 3) | SUM 9.83738s | PER ITER avg 0.09837s, min 0.09417s, max 0.10458s
|
|
[Augmenter: AffineOrder0]
|
|
(16, 4, 4, 3) | SUM 0.37980s | PER ITER avg 0.00380s, min 0.00340s, max 0.00517s
|
|
(16, 32, 32, 3) | SUM 0.53106s | PER ITER avg 0.00531s, min 0.00472s, max 0.00630s
|
|
(16, 256, 256, 3) | SUM 10.69961s | PER ITER avg 0.10700s, min 0.10223s, max 0.11325s
|
|
[Augmenter: AffineOrder1]
|
|
(16, 4, 4, 3) | SUM 0.39431s | PER ITER avg 0.00394s, min 0.00363s, max 0.00511s
|
|
(16, 32, 32, 3) | SUM 0.62730s | PER ITER avg 0.00627s, min 0.00576s, max 0.00711s
|
|
(16, 256, 256, 3) | SUM 14.50003s | PER ITER avg 0.14500s, min 0.13785s, max 0.15291s
|
|
[Augmenter: AffineAll]
|
|
(16, 4, 4, 3) | SUM 0.58742s | PER ITER avg 0.00587s, min 0.00429s, max 0.00724s
|
|
(16, 32, 32, 3) | SUM 3.69956s | PER ITER avg 0.03700s, min 0.01358s, max 0.06233s
|
|
(16, 256, 256, 3) | SUM 212.91776s | PER ITER avg 2.12918s, min 0.57114s, max 3.95389s
|
|
"""
|
|
|
|
|
|
def main():
|
|
augmenters = [
|
|
iaa.Identity(name="Identity"),
|
|
iaa.Crop(px=(0, 8), name="Crop-px"),
|
|
iaa.Crop(percent=(0, 0.1), name="Crop-percent"),
|
|
iaa.Fliplr(0.5, name="Fliplr"),
|
|
iaa.Flipud(0.5, name="Flipud"),
|
|
iaa.Grayscale((0.0, 1.0), name="Grayscale"),
|
|
iaa.GaussianBlur((0, 3.0), name="GaussianBlur"),
|
|
iaa.AdditiveGaussianNoise(loc=0, scale=(0.0, 0.1), name="AdditiveGaussianNoise"),
|
|
iaa.Dropout((0.0, 0.1), name="Dropout"),
|
|
iaa.Multiply((0.5, 1.5), name="Multiply"),
|
|
iaa.ContrastNormalization(alpha=(0.5, 2.0), name="ContrastNormalization"),
|
|
iaa.Grayscale(alpha=(0.0, 1.0), name="Grayscale"),
|
|
iaa.ElasticTransformation(alpha=(0.5, 8.0), sigma=1.0, name="ElasticTransformation"),
|
|
iaa.Affine(
|
|
scale={"x": (0.8, 1.2), "y": (0.8, 1.2)},
|
|
translate_px={"x": (-16, 16), "y": (-16, 16)},
|
|
rotate=(-45, 45),
|
|
shear=(-16, 16),
|
|
order=0,
|
|
cval=(0, 255),
|
|
mode="constant",
|
|
name="AffineOrder0ModeConstant"
|
|
)
|
|
]
|
|
|
|
for order in [0, 1]:
|
|
augmenters.append(
|
|
iaa.Affine(
|
|
scale={"x": (0.8, 1.2), "y": (0.8, 1.2)},
|
|
translate_px={"x": (-16, 16), "y": (-16, 16)},
|
|
rotate=(-45, 45),
|
|
shear=(-16, 16),
|
|
order=order,
|
|
cval=(0, 255),
|
|
mode=ia.ALL,
|
|
name="AffineOrder%d" % (order,)
|
|
)
|
|
)
|
|
|
|
augmenters.append(
|
|
iaa.Affine(
|
|
scale={"x": (0.8, 1.2), "y": (0.8, 1.2)},
|
|
translate_px={"x": (-16, 16), "y": (-16, 16)},
|
|
rotate=(-45, 45),
|
|
shear=(-16, 16),
|
|
order=ia.ALL,
|
|
cval=(0, 255),
|
|
mode=ia.ALL,
|
|
name="AffineAll"
|
|
)
|
|
)
|
|
|
|
kps = []
|
|
for _ in sm.xrange(20):
|
|
x = random.randint(0, 31)
|
|
y = random.randint(0, 31)
|
|
kps.append(ia.Keypoint(x=x, y=y))
|
|
kps = ia.KeypointsOnImage(kps, shape=(32, 32, 3))
|
|
small_keypoints_one = kps.on((4, 4, 3))
|
|
medium_keypoints_one = kps.on((32, 32, 3))
|
|
large_keypoints_one = kps.on((256, 256, 3))
|
|
small_keypoints = [small_keypoints_one.deepcopy() for _ in sm.xrange(16)]
|
|
medium_keypoints = [medium_keypoints_one.deepcopy() for _ in sm.xrange(16)]
|
|
large_keypoints = [large_keypoints_one.deepcopy() for _ in sm.xrange(16)]
|
|
|
|
small_images = np.random.randint(0, 255, (16, 4, 4, 3)).astype(np.uint8)
|
|
medium_images = np.random.randint(0, 255, (16, 32, 32, 3)).astype(np.uint8)
|
|
large_images = np.random.randint(0, 255, (16, 256, 256, 3)).astype(np.uint8)
|
|
|
|
print("---------------------------")
|
|
print("Keypoints")
|
|
print("---------------------------")
|
|
for augmenter in augmenters:
|
|
print("[Augmenter: %s]" % (augmenter.name,))
|
|
for keypoints in [small_keypoints, medium_keypoints, large_keypoints]:
|
|
times = []
|
|
for i in sm.xrange(100):
|
|
time_start = time.time()
|
|
_img_aug = augmenter.augment_keypoints(keypoints)
|
|
time_end = time.time()
|
|
times.append(time_end - time_start)
|
|
times = np.array(times)
|
|
img_str = "{:20s}".format(keypoints[0].shape)
|
|
print("%s | SUM %.5fs | PER ITER avg %.5fs, min %.5fs, max %.5fs" % (
|
|
img_str, float(np.sum(times)), np.average(times), np.min(times), np.max(times)))
|
|
|
|
print("---------------------------")
|
|
print("Images")
|
|
print("---------------------------")
|
|
for augmenter in augmenters:
|
|
print("[Augmenter: %s]" % (augmenter.name,))
|
|
for images in [small_images, medium_images, large_images]:
|
|
times = []
|
|
for i in sm.xrange(100):
|
|
time_start = time.time()
|
|
_img_aug = augmenter.augment_images(images)
|
|
time_end = time.time()
|
|
times.append(time_end - time_start)
|
|
times = np.array(times)
|
|
img_str = "{:20s}".format(images.shape)
|
|
print("%s | SUM %.5fs | PER ITER avg %.5fs, min %.5fs, max %.5fs" % (
|
|
img_str, float(np.sum(times)), np.average(times), np.min(times), np.max(times)))
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|