Tengine for Rasberry Pi(Education Edition)

For education only!

Tengine, developed by OPEN AI LAB, is a lite, high-performance, and modular inference engine for embedded device.
Github: https://github.com/OAID/Tengine

Installation

1. Install tools & libraries

   sudo apt-get install git cmake
   
   sudo apt-get install libprotobuf-dev protobuf-compiler
   sudo apt-get install libboost-all-dev
   sudo apt-get install libgoogle-glog-dev
   sudo apt-get install libopencv-dev
   sudo apt-get install libopenblas-dev

2. Get Tengine code

   git clone https://github.com/OAID/Tengine.git

3. Prepare config files

   • copy config example file

     cd ~/Tengine
     cp makefile.config.example makefile.config

   • edit makefile.config
     • comment CONFIG_ARCH_ARM64 option to invalid arm64 arch.
     • uncomment CONFIG_ARCH_BLAS=y option to valid Openblas arch.

4. Build

   make -j4

5. Get the binary library
   Replace Tengine/install/lib/libtengine.so with the new one.

Example

Reference: 树莓派也能玩转深度学习——Tengine推断引擎/测试 | Hey~YaHei!

With Tengine for Rasberry Pi(Education Edition), SPF(Second Per Frame) of mobilenet-ssd model drops from 1.148 to 0.286.