SDK Design

How to Design a Language Agnostic SDK for Cross Platform Deployment and Maximum Extensibility
A Venice Computer Vision Presentation

Video Presentation · Presentation Slides · Blog Post · Venice Computer Vision

SDK Design

This project demonstrates how to build a language agnostic SDK for cross platform deployment and maximum extensibility. It covers how to do the following:

• Build a basic face detection computer vision library in C++
• Compile / cross compile the library for amd64, arm64, arm32
• Package that library and its dependencies as a single static library
• Add unit tests
• Set up a CI pipeline
• Write python bindings for our library
• Generate documentation directly from our API

Please refer to the blog post for a detailed tutorial and explanation of all the components of this project.

Getting Started

The following instructions assume you are using Ubuntu 18.04

Prerequisites

• sudo apt install build-essential
• sudo apt-get install g++-aarch64-linux-gnu
• sudo apt-get install gcc-arm-linux-gnueabihf binutils-arm-linux-gnueabihf g++-arm-linux-gnueabihf
• sudo apt install python3.8
• sudo apt install python3-pip
• pip3 install cmake
• sudo apt-get install doxygen
• sudo apt-get install wget
• sudo apt-get install zip

Install 3rd Party Libraries

Navigate to 3rdparty then run the following:

• ./build_catch.sh
• ./build_pybind11.sh
• ./build_ncnn.sh
• ./build_opencv.sh

Building the library

• mkdir build && cd build
• cmake ..
• make -j$(nproc)
• make install

The outputs will be copied to dist

Cross comping for arm32

• mkdir build && cd build
• cmake -D BUILD_ARM32=ON ..
• make -j$(nproc)
• make install

Cross compiling for arm64

• mkdir build && cd build
• cmake -D BUILD_ARM64=ON ..
• make -j$(nproc)
• make install