# Copyright (C) 2019, STMicroelectronics - All Rights Reserved SUMMARY = "Create package containing MobileNetV1/2 models" LICENSE = "Apache-2.0" LIC_FILES_CHKSUM = "file://${COMMON_LICENSE_DIR}/Apache-2.0;md5=89aea4e17d99a7cacdbeed46a0096b10" # These models are not available in the onnx/models repository and integrating the tf2onnx # converter in the yocto workflow (to convert them from tflite) would require integrating # tensorflow (not lite) and bazel as well, which is a pretty big undertaking. # So, for now, we will simply include the converted models manually. SRC_URI = " file://mobilenet_v1_0.5_128.onnx \ file://mobilenet_v1_0.5_128_quant.onnx \ file://mobilenet_v1_1.0_224_quant.onnx \ file://labels_mobilenet_onnx.txt " S = "${WORKDIR}" do_configure[noexec] = "1" do_compile[noexec] = "1" do_install() { install -d ${D}${prefix}/local/demo-ai/image-classification/models/mobilenet install -d ${D}${prefix}/local/demo-ai/image-classification/models/mobilenet/testdata # install mobilenet models install -m 0644 ${S}/label*.txt ${D}${prefix}/local/demo-ai/image-classification/models/mobilenet/labels_onnx.txt install -m 0644 ${S}/*.onnx ${D}${prefix}/local/demo-ai/image-classification/models/mobilenet/ } FILES:${PN} += "${prefix}/local/"