Image Benchmark Results

Single-threaded CPU benchmark comparing image augmentation throughput (images/second). All libraries run under identical conditions on the same machine.

95 of 101
transforms where Albumentationsx is fastest
15.3x
average speedup over nearest competitor
across transforms where Albumentationsx leads
264x
fastest speedup (MedianBlur)
vs next fastest library

Pairwise Summary

Generated automatically from benchmark JSON at build time (median speedup midpoint with IQR).

Albumentationsx vs Kornia

Wins: 46 / 48 · Median: 6.32x · IQR: 2.65x-11.50x

Albumentationsx vs Torchvision

Wins: 24 / 27 · Median: 5.78x · IQR: 1.77x-10.27x

Speedup Distribution

How many transforms fall into each speedup range (Albumentationsx vs best competitor). 52 transforms with head-to-head comparison.

1< 0.5×40.5–1×111–2×122–5×155–10×610–50×3> 50×

Results

Transform
albumentationsx
2.1.1
CPU · macOS arm64
kornia
0.8.2
CPU · macOS arm64
torchvision
0.25.0
CPU · macOS arm64
Speedup
Albx / best other (range)
MedianBlur1550 ± 196 ± 0256-272x
Elastic425 ± 81 ± 03 ± 0115-132x
RandomGamma14281 ± 1142226 ± 557-70x
MotionBlur4265 ± 48117 ± 634-39x
Blur7402 ± 234365 ± 819-21x
Hue2001 ± 9123 ± 715-17x
PhotoMetricDistort1070 ± 880 ± 313-14x
Solarize13474 ± 919262 ± 31117 ± 3511-13x
ColorJitter1199 ± 25100 ± 388 ± 311-13x
Saturation1307 ± 38132 ± 49.4-11x
Rotate2914 ± 21330 ± 7319 ± 88.6-9.1x
Pad38630 ± 6734480 ± 1298.2-9.0x
LongestMaxSize3908 ± 47481 ± 367.5-8.9x
Grayscale17827 ± 11011574 ± 772206 ± 1797.0-9.3x
Sharpen2226 ± 44263 ± 14274 ± 97.7-8.5x
SmallestMaxSize2728 ± 24375 ± 107.0-7.5x
GaussianBlur2344 ± 46353 ± 13124 ± 176.3-7.0x
CLAHE652 ± 2109 ± 25.8-6.1x
HorizontalFlip12908 ± 4221128 ± 422234 ± 275.5-6.0x
Snow707 ± 11129 ± 45.2-5.7x
Affine1423 ± 7264 ± 165.1-5.7x
Perspective1167 ± 14170 ± 5217 ± 85.1-5.6x
Resize3477 ± 30648 ± 15271 ± 45.2-5.5x
RandomResizedCrop4226 ± 17579 ± 6789 ± 275.2-5.6x
Contrast9695 ± 2762159 ± 193870 ± 264.0-5.1x
Brightness9963 ± 3392276 ± 1691681 ± 213.9-4.9x
OpticalDistortion774 ± 13193 ± 43.9-4.2x
ChannelDropout11932 ± 4413065 ± 1793.5-4.3x
Shear1266 ± 10358 ± 113.4-3.7x
AutoContrast1670 ± 15576 ± 18178 ± 22.8-3.0x
RGBShift4736 ± 1231787 ± 712.5-2.8x
GaussianNoise319 ± 8121 ± 22.5-2.7x
Erasing26700 ± 3002776 ± 4510421 ± 6292.1-3.0x
PlasmaBrightness186 ± 176 ± 22.4-2.5x
Equalize1273 ± 6310 ± 17588 ± 172.1-2.2x
PlankianJitter3215 ± 251578 ± 1001.9-2.2x
ChannelShuffle7779 ± 521446 ± 1154290 ± 3031.7-2.0x
GaussianIllumination765 ± 11428 ± 161.7-1.9x
PlasmaContrast134 ± 475 ± 61.6-2.0x
JpegCompression1353 ± 1117 ± 5826 ± 111.6-1.7x
Invert33252 ± 31854412 ± 29322891 ± 24841.2-1.8x
CornerIllumination500 ± 3350 ± 41.4-1.5x
Normalize1621 ± 101173 ± 39947 ± 331.3-1.4x
ThinPlateSpline84 ± 561 ± 21.3-1.5x
Rain2174 ± 371591 ± 611.3-1.4x
SaltAndPepper599 ± 15450 ± 51.3-1.4x
VerticalFlip26665 ± 18512387 ± 5826928 ± 47990.8-1.3x
PlasmaShadow205 ± 7211 ± 50.9-1.0x
Posterize13288 ± 667709 ± 2717723 ± 13800.7-0.9x
LinearIllumination568 ± 5849 ± 220.6-0.7x
RandomCrop12864659 ± 11752802 ± 40112838 ± 23840.6-0.6x
CenterCrop12865127 ± 1358203348 ± 74290.3-0.3x
AdditiveNoise267 ± 12
AdvancedBlur1256 ± 49
AtmosphericFog349 ± 15
ChannelSwap7363 ± 76
ChromaticAberration496 ± 13
CoarseDropout18413 ± 1356
ConstrainedCoarseDropout108398 ± 1116
CropAndPad2763 ± 49
Defocus131 ± 1
Dithering6 ± 0
Downscale5351 ± 68
Emboss2707 ± 27
FancyPCA111 ± 2
FilmGrain272 ± 1
GlassBlur36 ± 1
GridDistortion1309 ± 38
GridDropout82 ± 3
GridMask15684 ± 535
HSV1140 ± 13
Halftone26 ± 1
ISONoise178 ± 3
LensFlare263 ± 10
Morphological17026 ± 1242
MultiplicativeNoise4955 ± 109
PadIfNeeded16014 ± 370
PiecewiseAffine173 ± 7
PixelDropout403 ± 9
RandomFog9 ± 0
RandomGravel1299 ± 11
RandomGridShuffle10619 ± 72
RandomRotate901866 ± 47
RandomScale3300 ± 44
RandomShadow507 ± 8
RandomSizedCrop3676 ± 66
RandomSunFlare347 ± 2
RandomToneCurve10725 ± 377
RingingOvershoot151 ± 3
SafeRotate1346 ± 4
ShiftScaleRotate1362 ± 18
ShotNoise45 ± 1
Spatter108 ± 3
SquareSymmetry2052 ± 46
Superpixels18 ± 0
ToSepia6866 ± 214
Transpose1580 ± 20
UnsharpMask383 ± 16
Vignetting589 ± 16
WaterRefraction124 ± 5
ZoomBlur197 ± 12

Methodology

Test Environment

Platform
macOS arm64 (Apple M-series)
CPU threads
1 (forced single-thread)
Images per run
2000
Runs per transform
5
Last run
March 27, 2026

Library Versions

Albumentationsx
2.1.1
Kornia
0.8.2
Torchvision
0.25.0
NumPy
2.4.3
OpenCV
4.13.0.92

Metric: Median throughput in images/second across 5 runs. Higher is better.

Warmup: Adaptive warmup until variance stabilizes before measurement begins.

Thread control: OMP, OpenBLAS, MKL, and OpenCV threads all forced to 1 to ensure fair single-thread comparison.

Image loading: Each library uses its native format — OpenCV (BGR→RGB) for Albumentationsx, normalized tensors for Kornia, PIL for Torchvision.

Speedup column: Range computed from throughput error bars: speedup_from to speedup_to, where each bound uses median +/- std. Green = 2x+ conservative bound, yellow = 1x-2x, gray = slower.

Want to verify the results on your own hardware or check that the comparison is fair? The benchmark code is open source on GitHub.