Skip to content
Accelerated Computer-vision Library

ACL Pack

Ultimate Speed, Pure Simplicity

Built with 163 NEON-optimized operators, 72% of operator samples faster than OpenCV with peak speedup 78×. With zero external dependencies, it is a top-tier image processing SDK for ARM platforms.

It efficiently unlocks your device's computing power, while drastically reducing memory footprint and processor cost. Affordably priced with exceptional value.
Offering Personal, Pro, Small Team, and Enterprise editions to fit teams of every scale.
ACL Pack high-performance ARM64 computer vision SDK

Why ACL Pack

ACL Pack SDK Key Features

163
Operators
9 categories covering filter, color, geometric, feature detection, and more
72.3%
Faster than OpenCV
Across 6,864 NEON operator samples, 72.3% are faster than OpenCV 4.13.0, with an aggregate speedup of 4.72×
78×
Peak Speedup
M tier peak 78× on Snapdragon 8 Gen 3; per-pixel ops up to 76×
99%
Save Engineering Cost
One license replaces 2–3 engineer-years of work — cheaper than in-house or AI-assisted development.

Fast & Memory-efficient

Performance

Heavy workloads drop from seconds to milliseconds, M tier peak 78×. In-place processing and zero-copy design minimize memory footprint on resource-constrained devices.

SIMD Hand-tuned

Architecture

Every core operator hand-tuned in NEON; open architecture — will extend to SVE / CME / OpenCL / HVX and more instruction sets.

Zero Dependencies

Lightweight

Pure C++ static library, no OpenCV, no Boost, no third-party runtime needed.

Multi-type & Format

Format

u8/u16/s16/f32 pixel types with full YUV, Bayer and RGB family support. All industry-standard formats, ready for commercial use.

Multi-ARM & Production Proven

Reliability

Runs across multiple ARM platforms — validated by 120,000+ cross-platform bit-identical tests (cross-device, cross-platform, multi-size).

License Protection

Security

RSA-2048 signed license, `licenseId` bound to order for traceability, one license per order with unlimited devices and no app-binding.

ACL Pack vs SOTA Competitor — Real Benchmarks

1920×1280, cross-device NEON peak

OperatorSpeedupACL vs SOTA
Resize NN Up 4×up to 78.22×
inRange (Mask)up to 76.42×
Threshold Binaryup to 61.79×
Resize AREA Up 4×up to 49.51×
BitwiseAndup to 29.59×
Resize AREA Up 2×up to 28.03×
Sobel gradXup to 26.03×
Alpha Fusion (α=0.5)up to 22.20×
Gaussian Blur 5×5up to 15.71×
Median Filter 3×3up to 14.99×
normalize MINMAXup to 14.39×
Rotate 180°up to 13.98×
BGR → Labup to 12.86×

Based on 1,318 cross-device benchmarks at 1920×1280 across 4 ARM64 devices.

Simple to Integrate

Include the header, link the library, call the function. That's it.

main.cpp
#include <acl/acl.h>
#include <acl/api.h>

// One-time license init, then call any operator. Zero external deps.
acl::init("license.dat");
acl::gaussianBlur(src, dst, w, h, 1, 0, 0, 1, 1);
acl::threshold   (src, dst, w, h, 128);
acl::resize      (src, dst, sw, sh, dw, dh);
4
Lines to start
0
External dependencies
≤ 3.7 MB
Library size

163 Operators Across 9 Groups

Both NEON-optimized and portable C++ implementations. Include only what you need. Continuously updated.

~

Filter

Linear/non-linear filtering, edge detection, morphology, denoising and HDR fusion.

31 operators
🎨

Color Conversion

RGB/BGR, grayscale, HSV, Lab, Bayer, YUV conversions and Gamma.

36 operators

Geometric

Resize, rotate, flip, image pyramids and YUV-aware geometric transforms.

11 operators
±

Arithmetic

Add/sub/mul, thresholding, bitwise ops, LUT, normalization, blending and range ops.

20 operators
📊

Analysis

Histograms, CLAHE, template matching, contours, connected components, distance transform and statistics.

20 operators
📍

Feature Detection

Corners, ORB/SIFT/SURF, HOG, Hough transform, optical flow and feature matching.

22 operators
📐

Transform

Affine/perspective warping, remap, homography estimation and error metrics.

10 operators

Math Transforms

1D/2D DFT, real-input DFT/IDFT and spectrum multiplication.

5 operators

Utilities

Drawing, contour analysis, convex hull, polygon approximation and ROI cropping.

12 operators