NobuoTsukamoto/meta-tensorflow-lite

Yocto layer for TensorFlow Lite interpreter with Python / C++.

meta-tensorflow-lite

Yocto layer for the TensorFlow Lite interpreter with Python / C++.

The official website is:

• TensorFlow Lite guide
• Python quickstart
• TensorFlow Lite
• Google Coral
• Edge TPU runtime library (libedgetpu)

Reference

• Building TensorFlow Lite Standalone Pip
• Build TensorFlow Lite with CMake
• Building with only a Makefile - libedgetpu

Available BSP

Please note that it is not official support.

BSP	Build status
meta-raspberrypi
meta-riscv

Available recipes

• framework
  • python3-tensorflow-lite
    Python3 interpreter
  • libtensorflow-lite
    C++ API shared library
  • libtensorflow-lite-c
    C API shared library
  • libedgetpu-max / libedgetpu-std
    bitbake with libedgetpu
• examples
  • python3-tensorflow-lite-example
    TensorFlow Lite Python image classification demo
  • tensorflow-lite-label-image
    TensorFlow Lite C++ image classification demo
  • tensorflow-lite-minimal
    TensorFlow Lite C++ minimal example
• tools
  • tensorflow-lite-benchmark
    TFLite Model Benchmark Tool with C++ Binary

How to

Quick start for the Raspberry Pi AArch64 (core-image-weston)

# Clone repositories and oe-init-build-env
$ git clone git://git.yoctoproject.org/poky.git
$ git clone git://git.yoctoproject.org/meta-raspberrypi
$ git clone git://git.openembedded.org/meta-openembedded
$ git clone https://github.com/NobuoTsukamoto/meta-tensorflow-lite.git
$ source poky/oe-init-build-env build

# Add layer
$ bitbake-layers add-layer ../meta-openembedded/meta-oe/
$ bitbake-layers add-layer ../meta-openembedded/meta-python/
$ bitbake-layers add-layer ../meta-openembedded/meta-networking/
$ bitbake-layers add-layer ../meta-openembedded/meta-multimedia/
$ bitbake-layers add-layer ../meta-raspberrypi/
$ bitbake-layers add-layer ../meta-tensorflow-lite/

# Add the package to 'conf/auto.conf' file. 
FORTRAN:forcevariable = ",fortran"
MACHINE ?= "raspberrypi4-64"
IMAGE_INSTALL:append = " python3-tensorflow-lite libtensorflow-lite"

# Build
$ bitbake core-image-weston

Quick start for the qemuriscv64

# Clone repositories and oe-init-build-env
$ git clone https://github.com/openembedded/bitbake.git
$ git clone https://github.com/openembedded/openembedded-core.git
$ git clone https://github.com/openembedded/meta-openembedded.git
$ git clone https://github.com/riscv/meta-riscv.git
$ git clone https://github.com/NobuoTsukamoto/meta-tensorflow-lite.git
$ source openembedded-core/oe-init-build-env build

# Add layer
$ bitbake-layers add-layer ../meta-openembedded/meta-oe/
$ bitbake-layers add-layer ../meta-openembedded/meta-python/
$ bitbake-layers add-layer ../meta-openembedded/meta-networking/
$ bitbake-layers add-layer ../meta-openembedded/meta-multimedia/
$ bitbake-layers add-layer ../meta-riscv/
$ bitbake-layers add-layer ../meta-tensorflow-lite/

# Add the package to 'conf/auto.conf' file. 
FORTRAN:forcevariable = ",fortran"
MACHINE ?= "qemuriscv64"
IMAGE_INSTALL:append = " python3-tensorflow-lite  libtensorflow-lite"

# Build
$ bitbake core-image-full-cmdline