English | 简体中文
Mini runtime that suitable for model inference in server. Current packages build core, imgproc, imgcodecs, and, since 4.13.0.45, dnn.
osx-arm64 is the only exception: it still ships core, imgproc, and imgcodecs, but excludes dnn because GitHub Actions macOS arm64 runners repeatedly stall during OpenCV DNN builds.
Packages are all built & integration-tested by a fully automated CI/CD pipeline on GitHub, CICD code fully open source.
| OS | Package Id | NuGet | Compiler | Mini-ABI | Modules |
|---|---|---|---|---|---|
| Windows | Sdcb.OpenCvSharp4.mini.runtime.win-x64 |  |
MSVC 17 | win-x64 | core, imgproc, imgcodecs, dnn |
| Sdcb.OpenCvSharp4.mini.runtime.win-x86 |  |
MSVC 17 | win-x86 | core, imgproc, imgcodecs, dnn | |
| Sdcb.OpenCvSharp4.mini.runtime.win-arm64 |  |
MSVC 17 | win-arm64 | core, imgproc, imgcodecs, dnn | |
| Linux (glibc) | Sdcb.OpenCvSharp4.mini.runtime.linux-x64 |  |
GCC 8.5.0 | glibc 2.28+ | core, imgproc, imgcodecs, dnn |
| Sdcb.OpenCvSharp4.mini.runtime.linux-arm64 |  |
GCC 8.5.0 | glibc 2.28+ | core, imgproc, imgcodecs, dnn | |
| Linux (musl) | Sdcb.OpenCvSharp4.mini.runtime.linux-musl-x64 |  |
GCC 14.2.0 | musl 1.2 static | core, imgproc, imgcodecs, dnn |
| Sdcb.OpenCvSharp4.mini.runtime.linux-musl-arm64 |  |
GCC 14.2.0 | musl 1.2 static | core, imgproc, imgcodecs, dnn | |
| macOS | Sdcb.OpenCvSharp4.mini.runtime.osx-x64 |  |
Xcode 16 | macOS 10.15+ | core, imgproc, imgcodecs, dnn |
| Sdcb.OpenCvSharp4.mini.runtime.osx-arm64 |  |
Xcode 16 | macOS 11+ | core, imgproc, imgcodecs | |
| Android | Sdcb.OpenCvSharp4.mini.runtime.android-x64 |  |
Clang (NDK r27) | API 24+, static | core, imgproc, imgcodecs, dnn |
| Sdcb.OpenCvSharp4.mini.runtime.android-arm64 |  |
Clang (NDK r27) | API 24+, static | core, imgproc, imgcodecs, dnn |
These packages remain on NuGet for existing users, but are no longer actively built or validated by the current pipeline.
| OS / Baseline | Package Id Prefix | Last Intended Use | Recommended Replacement |
|---|---|---|---|
| CentOS 7 | Sdcb.OpenCvSharp4.mini.runtime.centos.7-* | glibc 2.17 era systems | linux-x64 / linux-arm64 |
| Ubuntu 22.04 | Sdcb.OpenCvSharp4.mini.runtime.ubuntu.22.04-* | glibc 2.35-specific builds | linux-x64 / linux-arm64 |
| Ubuntu 24.04 | Sdcb.OpenCvSharp4.mini.runtime.ubuntu.24.04-* | experimental newer-glibc builds | linux-x64 / linux-arm64 |
| Windows 11 | Sdcb.OpenCvSharp4.mini.runtime.win11-x64 | Windows 11-specific label experiment | win-x64 |
| macOS 15 | Sdcb.OpenCvSharp4.mini.runtime.osx.15-arm64 | macOS 15-specific label experiment | osx-arm64 |
| RHEL 9 | Sdcb.OpenCvSharp4.mini.runtime.rhel9-* | planned but not maintained separately | linux-x64 / linux-arm64 |
- 11 package build/link smoke tests
- 11 native host runtime tests
- 36 Docker container compatibility tests
- Android-x64 emulator runtime test
Assurance: every package runs reliably on its target platform.
• Use win-x64 (most PCs) or win-x86 (legacy).
• Tested on Windows Server 2022 & 2025; older versions should work thanks to Windows’ strong ABI compatibility.
• win-arm64 is now available and verified for Windows on ARM devices.
The maintained linux-x64 / arm64 packages are built on the RHEL 8 baseline and dynamically link libstdc++.so.6 + libgcc_s.so.1.
They run on glibc 2.28+ distributions, including RHEL 8+, Debian 10+, Ubuntu 20.04+, openEuler 20.03-lts, Deepin, and newer systems.
Packages linux-musl-x64 / arm64 are built on Alpine 3.22 and statically link libstdc++ & libgcc.
Confirmed to run on Alpine 3.12 (musl 1.1.24) — even routers running OpenWrt!
android-x64 / arm64 compiled with Android NDK r27, API 24.
libstdc++ is statically linked, so drop-in ready for any device running Android 7.0+.
• osx-x64 for Intel Macs.
• osx-arm64 for Apple Silicon (M1 → M4).
Both are tested on macOS 15 runners. osx-arm64 currently excludes dnn because OpenCV DNN builds repeatedly stall on GitHub Actions macOS arm64 infrastructure.
If these runtimes make your life easier, please…
- Star the repo: https://github.com/sdcb/opencvsharp-mini-runtime
- Consider a small donation – let me know your special needs and I’ll try to publish a build that fits.
Thanks for using OpenCvSharp mini-runtime and happy hacking!
— @Sdcb