mobileSSD 則是在 readme中說明步驟，原來的 model 是 caffe。
for Android 的部分:
yolonas 提供一個 script，download 需要的opencv source和把convert 好的dlc 和 spne 的 so copy 到 source folder.
mobileSSD 在 reame 中有說明步驟。
Mobilenet SSD
mobilenetssd 是用 caffe，所以snpe 要裝好 caffe。
snpe 中各個framework 的 tool 都是獨立的。所以可以針對tool 開conda env.
我用 ubuntu18.04，用update-alternatives把系統 python 改成 python3 (3.6) 來做。
snpe 1.68 的 caffe converting tool : snpe-caffe-to-dlc 是 python script，script head 指定用 python3。
所以build caffe 時要設定 support python3。
依照build caffe for snpe，build 好標準版 caffe。
依照 mobilessd 的說明 download prototxt 跟 caffemodel，用 snpe-caffe-to-dlc 做轉換，出現 Error:
l$ snpe-caffe-to-dlc --input_network deploy.prototxt --caffe_bin mobilenet_iter_73000.caffemodel --output_path moblie_net.dlc ERROR_CAFFE_CAFFE_PARSING_ERROR: Caffe could not parse deploy.prototxt: 2367:3 : Message type "caffe.LayerParameter" has no field named "permute_param". INFO_CAFFE_CAFFE_INSTALLATION_ERROR: Caffe installation in use: /home/charles-chang/caffe/python/caffe/__init__.py說caffe 不認識 permute_param 這個 layer。
應該要用caffe ssd這個版本的 caffe。
所以依照說說明，使用這個版本，checkout ssd，並且build for python3，修改的 Makefile:
$ git diff
diff --git a/Makefile b/Makefile
index 3fd68d1d..0789ae64 100644
--- a/Makefile
+++ b/Makefile
@@ -34,7 +34,7 @@ LIB_BUILD_DIR := $(BUILD_DIR)/lib
STATIC_NAME := $(LIB_BUILD_DIR)/lib$(LIBRARY_NAME).a
DYNAMIC_VERSION_MAJOR := 1
DYNAMIC_VERSION_MINOR := 0
-DYNAMIC_VERSION_REVISION := 0-rc3
+DYNAMIC_VERSION_REVISION := 0
DYNAMIC_NAME_SHORT := lib$(LIBRARY_NAME).so
#DYNAMIC_SONAME_SHORT := $(DYNAMIC_NAME_SHORT).$(DYNAMIC_VERSION_MAJOR)
DYNAMIC_VERSIONED_NAME_SHORT := $(DYNAMIC_NAME_SHORT).$(DYNAMIC_VERSION_MAJOR).$(DYNAMIC_VERSION_MINOR).$(DYNAMIC_VERSION_REVISION)
@@ -178,7 +178,7 @@ ifneq ($(CPU_ONLY), 1)
LIBRARIES := cudart cublas curand
endif
-LIBRARIES += glog gflags protobuf boost_system boost_filesystem boost_regex m hdf5_hl hdf5
+LIBRARIES += glog gflags protobuf boost_system boost_filesystem boost_regex m hdf5_serial_hl hdf5_serial
# handle IO dependencies
USE_LEVELDB ?= 1
@@ -328,6 +328,12 @@ ifeq ($(USE_CUDNN), 1)
COMMON_FLAGS += -DUSE_CUDNN
endif
+# NCCL acceleration configuration
+ifeq ($(USE_NCCL), 1)
+ LIBRARIES += nccl
+ COMMON_FLAGS += -DUSE_NCCL
+endif
+
# configure IO libraries
ifeq ($(USE_OPENCV), 1)
COMMON_FLAGS += -DUSE_OPENCV
@@ -571,7 +577,7 @@ $(STATIC_NAME): $(OBJS) | $(LIB_BUILD_DIR)
@ echo AR -o $@
$(Q)ar rcs $@ $(OBJS)
-$(BUILD_DIR)/%.o: %.cpp | $(ALL_BUILD_DIRS)
+$(BUILD_DIR)/%.o: %.cpp $(PROTO_GEN_HEADER) | $(ALL_BUILD_DIRS)
@ echo CXX $<
$(Q)$(CXX) $< $(CXXFLAGS) -c -o $@ 2> $@.$(WARNS_EXT) \
|| (cat $@.$(WARNS_EXT); exit 1)
@@ -688,6 +694,6 @@ $(DISTRIBUTE_DIR): all py | $(DISTRIBUTE_SUBDIRS)
install -m 644 $(DYNAMIC_NAME) $(DISTRIBUTE_DIR)/lib
cd $(DISTRIBUTE_DIR)/lib; rm -f $(DYNAMIC_NAME_SHORT); ln -s $(DYNAMIC_VERSIONED_NAME_SHORT) $(DYNAMIC_NAME_SHORT)
# add python - it's not the standard way, indeed...
- cp -r python $(DISTRIBUTE_DIR)/python
+ cp -r python $(DISTRIBUTE_DIR)/
-include $(DEPS)
Makefile.config.example (其實要 copy 成 Makefile.config
diff --git a/Makefile.config.example b/Makefile.config.example index eac93123..f82f01e4 100644 --- a/Makefile.config.example +++ b/Makefile.config.example @@ -31,21 +31,19 @@ CUDA_DIR := /usr/local/cuda # CUDA_DIR := /usr # CUDA architecture setting: going with all of them. -# For CUDA < 6.0, comment the lines after *_35 for compatibility. -CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \ - -gencode arch=compute_20,code=sm_21 \ - -gencode arch=compute_30,code=sm_30 \ - -gencode arch=compute_35,code=sm_35 \ - -gencode arch=compute_50,code=sm_50 \ - -gencode arch=compute_52,code=sm_52 \ - -gencode arch=compute_61,code=sm_61 +# For CUDA < 6.0, comment the *_50 through *_61 lines for compatibility. +# For CUDA < 8.0, comment the *_60 and *_61 lines for compatibility. +# For CUDA >= 9.0, comment the *_20 and *_21 lines for compatibility. +CUDA_ARCH := \ + -gencode arch=compute_60,code=sm_60 \ + -gencode arch=compute_61,code=sm_61 \ + -gencode arch=compute_61,code=compute_61 # BLAS choice: # atlas for ATLAS (default) # mkl for MKL # open for OpenBlas -# BLAS := atlas -BLAS := open +BLAS := atlas # Custom (MKL/ATLAS/OpenBLAS) include and lib directories. # Leave commented to accept the defaults for your choice of BLAS # (which should work)! @@ -63,19 +61,19 @@ BLAS := open # NOTE: this is required only if you will compile the python interface. # We need to be able to find Python.h and numpy/arrayobject.h. -PYTHON_INCLUDE := /usr/include/python2.7 \ - /usr/lib/python2.7/dist-packages/numpy/core/include +#PYTHON_INCLUDE := /usr/include/python2.7 \ +# /usr/lib/python2.7/dist-packages/numpy/core/include # Anaconda Python distribution is quite popular. Include path: # Verify anaconda location, sometimes it's in root. -# ANACONDA_HOME := $(HOME)/anaconda2 +# ANACONDA_HOME := $(HOME)/anaconda # PYTHON_INCLUDE := $(ANACONDA_HOME)/include \ - $(ANACONDA_HOME)/include/python2.7 \ - $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include \ + # $(ANACONDA_HOME)/include/python2.7 \ + # $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include # Uncomment to use Python 3 (default is Python 2) -# PYTHON_LIBRARIES := boost_python3 python3.5m -# PYTHON_INCLUDE := /usr/include/python3.5m \ -# /usr/lib/python3.5/dist-packages/numpy/core/include + PYTHON_LIBRARIES := boost_python3 python3.6m + PYTHON_INCLUDE := /usr/include/python3.6m \ + /usr/lib/python3.6/dist-packages/numpy/core/include # We need to be able to find libpythonX.X.so or .dylib. PYTHON_LIB := /usr/lib @@ -89,16 +87,20 @@ PYTHON_LIB := /usr/lib # WITH_PYTHON_LAYER := 1 # Whatever else you find you need goes here. -INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include +INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial/ LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib # If Homebrew is installed at a non standard location (for example your home directory) and you use it for general dependencies # INCLUDE_DIRS += $(shell brew --prefix)/include # LIBRARY_DIRS += $(shell brew --prefix)/lib +# NCCL acceleration switch (uncomment to build with NCCL) +# https://github.com/NVIDIA/nccl (last tested version: v1.2.3-1+cuda8.0) +# USE_NCCL := 1 + # Uncomment to use `pkg-config` to specify OpenCV library paths. # (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.) -# USE_PKG_CONFIG := 1 +USE_PKG_CONFIG := 1 # N.B. both build and distribute dirs are cleared on `make clean` BUILD_DIR := build
有關 snpe 做 MobiienetSSD 的，從 install(build caffe) ，train 到 convert to dlc 的步驟: 這個實際上是他門自己專案用的： 這個project clone 下來直接 build 就可以在 snapdragon 的手機上 run.
會先用 GPU 來 load model，失敗再用 CPU，在 pixel2 上 GPU 開啟失敗，只能用 CPU 啟動。
在 Qualcomm 的 QCS610 上，這個 snpe-release.aar 會 complain 不是 snapdragon，所以不 run
改用 snpe-1.63 版的 snpe-release.aar 和用這一版convert 的 dlc 來 build 就可以在 QCS610 上 run 了。
但是這樣在 pixel 2 上就 fail，說 permission 跟 so 有問題。
這個版本和qidk 的 mobilenetSSD 是用一樣的 model (prototxt, caffe bin)，但是 qidk 的 example run 起來是 fail 的。
用 android studio 看，這個 dashcam 跟 infer output 要的 key 是 detectio_out，但是 qidk 要的是 detection_output_number_detection，
在 android studio 中watch infer 後的 output，key 只有 detection_out，大概就是 qidk 的 example 沒有偵測出物件的原因吧？
這一篇問題的回覆可以看到 infer 出來的結果。
沒有留言:
張貼留言