hardware-neuralnetworks

Documentation

• The [release notes][] provide information on the release. [release notes]: https://wiki.st.com/stm32mpu/wiki/STM32_MPU_OpenSTDroid_release_note_-_v6.2.0

Prerequisites

The Android NDK must be installed (see : https://developer.android.com/ndk/downloads)

• Download and extract the Android NDK in your desired directory (here example ''$HOME/Android/''):
• export the ANDROID_NDK_TOOLCHAIN environment variable (can be added in your .bashrc)

export ANDROID_NDK_TOOLCHAIN="$HOME/Android/android-ndk-r29"

How to load sources (VSI NN libraries and tensorflow lite test binaries)

Ensure that you setup your Android distribution first, then execute:

load_nn

How to build loaded sources and install generated binaries (VSI NN libraries and tensorflow lite test binaries)

build_nn -i

Remark : it uses the default OpenVX SDK in prebuilt/android_arm64 subdirectory, it's compatible with gcnano 6.4.21 OpenVX.

How to test

VTS command

atest VtsHalNeuralnetworksTargetTest --test-filter=TestGenerated/GeneratedTest.Test/android_hardware_neuralnetworks_IDevice_nnapi_sample_sl_updatable_reshape

Test command

/vendor/bin/benchmark_model --graph=/vendor/etc/models/mobilenet_v1_1_0_224_quant_tflite --use_nnapi=true --nnapi_support_library_path=/vendor/lib64/libVsiSupportLibrary.so --nnapi_accelerator_name=vsi-device-0

License

This module is distributed under the proprietary license found in the LICENSE file.