Stay updated

News & Insights

Image Benchmark Results

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

50 of 54
transforms where Albumentationsx is fastest
14.6x
average speedup over nearest competitor
across transforms where Albumentationsx leads
261x
fastest speedup (MedianBlur)
vs next fastest library

Speedup Distribution

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

< 0.5×40.5–1×131–2×92–5×155–10×810–50×3> 50×

Results

Transform
albumentationsx
2.0.18
CPU · macOS arm64
kornia
0.8.2
CPU · macOS arm64
torchvision
0.25.0
CPU · macOS arm64
Speedup
Albx / best other
MedianBlur1533 ± 96 ± 0261x
Elastic362 ± 31 ± 03 ± 0105x
RandomGamma13280 ± 1279226 ± 559x
MotionBlur4385 ± 110117 ± 638x
Blur7592 ± 285365 ± 821x
Hue1917 ± 31123 ± 716x
PhotoMetricDistort943 ± 1880 ± 312x
Solarize12811 ± 785262 ± 31117 ± 3511x
ColorJitter1132 ± 23100 ± 388 ± 311x
Pad47542 ± 8204480 ± 12911x
Saturation1328 ± 45132 ± 410x
Grayscale20430 ± 22451574 ± 772206 ± 1799.3x
Rotate2981 ± 11330 ± 7319 ± 89.0x
Sharpen2251 ± 15263 ± 14274 ± 98.2x
LongestMaxSize3840 ± 68481 ± 368.0x
SmallestMaxSize2621 ± 31375 ± 107.0x
GaussianBlur2429 ± 9353 ± 13124 ± 176.9x
Contrast14165 ± 1042159 ± 193870 ± 266.6x
HorizontalFlip13654 ± 3531128 ± 422234 ± 276.1x
CLAHE633 ± 3109 ± 25.8x
Brightness12784 ± 10172276 ± 1691681 ± 215.6x
Snow723 ± 5129 ± 45.6x
RandomResizedCrop4322 ± 9579 ± 6789 ± 275.5x
Resize3502 ± 52648 ± 15271 ± 45.4x
Affine1428 ± 2264 ± 165.4x
Perspective1173 ± 3170 ± 5217 ± 85.4x
OpticalDistortion801 ± 2193 ± 44.1x
ChannelDropout12420 ± 8663065 ± 1794.1x
Shear1290 ± 9358 ± 113.6x
AutoContrast1666 ± 15576 ± 18178 ± 22.9x
GaussianNoise343 ± 4121 ± 22.8x
Erasing26411 ± 4926776 ± 4510421 ± 6292.5x
PlasmaBrightness170 ± 876 ± 22.2x
Equalize1243 ± 6310 ± 17588 ± 172.1x
PlasmaContrast156 ± 275 ± 62.1x
PlankianJitter3138 ± 691578 ± 1002.0x
ChannelShuffle8075 ± 2911446 ± 1154290 ± 3031.9x
GaussianIllumination772 ± 17428 ± 161.8x
JpegCompression1321 ± 9117 ± 5826 ± 111.6x
ThinPlateSpline89 ± 261 ± 21.5x
Invert32495 ± 63544412 ± 29322891 ± 24841.4x
Normalize1602 ± 91173 ± 39947 ± 331.4x
SaltAndPepper613 ± 4450 ± 51.4x
CornerIllumination468 ± 11350 ± 41.3x
Rain2064 ± 151591 ± 611.3x
RGBShift2252 ± 231787 ± 711.3x
VerticalFlip31055 ± 3252387 ± 5826928 ± 47991.2x
RandomCrop128113953 ± 27312802 ± 40112838 ± 23841.0x
PlasmaShadow196 ± 2211 ± 50.9x
Posterize13203 ± 680709 ± 2717723 ± 13800.7x
LinearIllumination485 ± 9849 ± 220.6x
CenterCrop128115895 ± 4274203348 ± 74290.6x
CoarseDropout23025 ± 2717
HSV1166 ± 20

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
February 26, 2026

Library Versions

Albumentationsx
2.0.18
Kornia
0.8.2
Torchvision
0.25.0
NumPy
2.4.2
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: Albumentationsx median ÷ highest competitor median. Green = 2× or faster, yellow = 1–2×, red = 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.