Tensorflow Lite Object Detection Library

Content

• Introduction
• Build Steps
  • On-device Compiling
  • Raspberry PI Cross Compiling

Introduction

This is a C++ library for tensorflow lite object detecion models.

The base ObjectDetector class is developed to use with ssd_mobilenet_v1 model.

To customize for other model, user need to derive ObjectDetector and rewrite PreProcess and PostProcess functions to align with the model inputs and outputs.

This library also has the configuration for setting up the model or labels path, and other parameter.

For usage, see the minimal example examples/minimal.

Build Steps

This library uses CMake to configurate and build. So you need to install CMake tool first.

Then clone the repository:

git clone --recursive https://github.com/bienxanh1901/tflite-object-detection

On-Device Compiling

• Update The system:

  sudo apt update
  sudo apt dist-upgrade

• Install development packages:

  sudo apt install build-essential cmake unzip pkg-config gfortran gcc g++ gperf flex texinfo gawk bison openssl pigz libncurses-dev autoconf automake tar figlet

• Build and install opencv:

  Follow this instruction to build and install opencv in folder thirparty/opencv: https://docs.opencv.org/4.x/d0/d3d/tutorial_general_install.html

• Create build directory:

  mkdir build && cd build

• Configurate:

  cmake ..

  or enable BUILD_EXAMPLES option to build examples:

  cmake -DBUILD_EXAMPLES ..

• Build and install

  cmake --build . -j <number-process>
  
  cmake --install . --prefix "<path/to/install/dir>"

Raspberry PI Cross Compiling

See cross-build-raspberry-PI.md