可以點下面連結去首頁
也可以直接點下面是 Tensorflow Lite 的資源
如下圖,目前是tensorflow2.6.0版本( dunfell )
如下圖,點擊 web repo
會到Github的頁面
git clone -b dunfell https://github.com/nnstreamer/meta-neural-network.git
裡面內容大概長成下面:
編譯一次nvt-image-qt5的Image , bitbake -s | grep tensorflow-lite 就應該可以看到tensorflow lite 的 library 被加到Image 裡面。
Debug 後 , 發現要 -DTFLITE_ENABLE_XNNPACK=off , 把XNNPACK給off掉,如下圖所示
EXTRA_OECMAKE = "\
-DTFLITE_ENABLE_XNNPACK=off \
-DTFLITE_ENABLE_RUY=on \
-DTFLITE_ENABLE_GPU=off \
${S}/tensorflow/lite/ \
下面這幾個library 是跟tigervnc相關:
gtk+3 、 tigervnc 、 xauth 、 xkbcomp 、 xsetroot 、 xterm 、 twm
MACHINE ??= 'numaker-som-ma35d16a81'
DISTRO ?= 'nvt-ma35d1-directfb'
PACKAGE_CLASSES ?= "package_rpm"
EXTRA_IMAGE_FEATURES ?= "debug-tweaks"
USER_CLASSES ?= "buildstats image-mklibs image-prelink"
PATCHRESOLVE = "noop"
BB_DISKMON_DIRS ??= "\
STOPTASKS,${TMPDIR},1G,100K \
STOPTASKS,${DL_DIR},1G,100K \
STOPTASKS,${SSTATE_DIR},1G,100K \
STOPTASKS,/tmp,100M,100K \
ABORT,${TMPDIR},100M,1K \
ABORT,${DL_DIR},100M,1K \
ABORT,${SSTATE_DIR},100M,1K \
ABORT,/tmp,10M,1K"
PACKAGECONFIG_append_pn-qemu-system-native = " sdl"
CONF_VERSION = "1"
MACHINE_FEATURES_append = " optee "
IMAGE_INSTALL_append += " opencv"
IMAGE_INSTALL_append += " tensorflow-lite"
IMAGE_INSTALL_append += " gtk+3"
IMAGE_INSTALL_append += " tigervnc"
IMAGE_INSTALL_append += " xauth"
IMAGE_INSTALL_append += " xkbcomp"
IMAGE_INSTALL_append += " xsetroot"
IMAGE_INSTALL_append += " xterm"
IMAGE_INSTALL_append += " twm"
IMAGE_INSTALL_append += " cmake"
DL_DIR ?= "${BSPDIR}/downloads/"
編譯 nvt-image-qt5 的 Image 時候會看到 下面 x11 錯誤
點開下面的 nvt-ma35d1-directfb.conf 檔案,加入x11 的內容就可以了
路徑:
/home/nuvoton/yocto/sources/meta-ma35d1/conf/distro
# Nuvoton DISTRO for the FrameBuffer graphical backend.
include conf/distro/include/nvt-base.inc
DISTRO = "nvt-ma35d1-directfb"
# Remove conflicting backends.
DISTRO_FEATURES_append += " "
DISTRO_FEATURES_append += " x11 "
再 bitbake nvt-image-qt5 一次,就可以全部都加進去