Guide Of Tool Chain Installation And Usage

User Manual:

Open the PDF directly: View PDF PDF.
Page Count: 6

Guide of Tool Chain Installation and Use
ver 1.0.0
content
1 Arm Cross Compile Tool Chain Installation ........................................ 2
1.1 Tool Chain Decompression ................................................................... 2
1.2 Tool Chain Installation .......................................................................... 2
1.3 Environment Variables Settings ............................................................ 2
1.4 Tool Chain Installation Test .................................................................. 3
2 Model Conversion Tool Installation ..................................................... 3
2.1 Tool Chain Decompression ................................................................... 3
2.2 Configuration File Description ............................................................. 4
2.3 Tool chain Installation .......................................................................... 4
2.4 Tool Chain Installation Test .................................................................. 4
3 Instructions for the use of tool chains ................................................... 5
3.1 Arm Compile Steps ............................................................................... 5
3.2 Model Conversion steps ....................................................................... 5
Release History
ver 1.0.0
2019/03/16
First version content
1 ARM CROSS COMPILE TOOL CHAIN INSTALLATION
This SDK provides the arm-hisiv300-linux cross compile toolchain. The installation
steps are as follows:
1.1 Tool Chain Decompression
The arm cross compile toolchain and its installer are located in directory
tools/arm_toolchain, enter this directory and decompress it. The commands are as
follows:
cd tools/arm_toolchain/
tar –xvf arm-hisi300-linux.tar.bz2
1.2 Tool Chain Installation
Enter the folder arm-hisi300-linux/ after it’s decompression and use the script to
install the toolchain. The commands are as follows:
chmod +x cross.v300.install
sudo ./ cross.v300.install
1.3 Environment Variables Settings
If your computer is 64-bit, you need to install 32-bit compiler compatibility
packages:
sudo aptitude install lib32z1
sudo aptitude install lib32stdc++6-4.8-dbg
Import the path of the tool chain into environment variable, the path can be viewed as
follows:
cd /opt/hisi-linux/x86-arm/arm-hisiv300-linux/target/bin
pwd
Add the following statement at the end of the file bashrc:
vi ~/.bashrc
export PATH=/opt/hisi-linux/x86-arm/arm-hisiv300-linux/target/bin:$PATH
Ensure that environmental variable settings take effect:
source ~/.bashrc
1.4 Tool Chain Installation Test
Execute the following commands to view the version of the toolchain:
arm-hisiv300-linux-gcc –v
If “gcc version 4.8.3 20131202 (prerelease) (Hisilicon_v300)” appears at the tail of
the version description, the toolchain installation is successful.
2 MODEL CONVERSION TOOL INSTALLATION
2.1 Tool Chain Decompression
The model conversion tool and its installer are located in directory
tools/mv_toolchain, enter this directory and decompress it. The commands are as follows:
cd tools/mv_toolchain/
tar –xvf model_conversion.tar.gz
2.2 Configuration File Description
Enter the folder after decompression and the development environment requirements
document is as follows:
── install-ncsdk.sh
#script to install all packeges and libs on your system, including caffe-ssd cpu
version and tensorflow.
── uninstall-ncsdk.sh # Uninstall
── ncsdk.conf # ncsdk config file.
── requirments.txt # python3 dependencies.
── requirments_apt.txt # ubuntu system dependencies.
2.3 Tool chain Installation
Use the script to install the toolchain. The commands are as follows:
sudo ./ install-ncsdk.sh
The installation log will be written into setup-logs.
2.4 Tool Chain Installation Test
Execute the following commands to view the version of the toolchain:
mvNCCompile -v
If “mvNCCompile v02.00, Copyright @ Movidius Ltd 2016” appears at the head of
the version description, the toolchain installation is successful. Use “-h” command to
see the help message:
mvNCCompile –h
3 INSTRUCTIONS FOR THE USE OF TOOL CHAINS
3.1 Arm Compile Steps
Developers need to use arm cross-compiler tool chain to compile their own projects,
generate executable files and run it on the embedded board. In this SDK sample program,
Makefile shows how to use arm cross compiler to generate executable files.
For example, enter folder examples/detect/, use “make” to compile and link, you will
see the compile steps like follows:
“ arm-hisiv300-linux-g++ -c xxx.c –o xxx.o
arm-hisiv300-linux-g++ -o xxx xxx.o –lpthread ../../libs/SIM_CAM_lib.a ”
xxx.c is user-developed programs, xxx.o is object-file, xxx is the final executable file.
3.2 Model Conversion steps
Developers need to use mvNCCompile tool chain to generate a model file which can
be read and run on Movidius. This model file could be named as “graph” and it contains
the structure and weight information of the neural network model trained by developers.
The often-used usage and parameters of mvNCCompile are as follows:
-h # Show help
-w WEIGHTS # Model weight (Caffe only), *.caffemodel*
-in INPUTNODE # Name of input node, default: Caffe: *data*, TF: *input*
-on OUTPUTNODE # Name of output node, default: Caffe: Last layer, TF:
-o OUTFILE # Output path of SimCam graph
-s NSHAVES # Number of Movidius Shave, recommend 6, default 1
For example, enter folder examples/convert/, use script compile.sh to compile, you
will see the compile steps like follows:
“ mvNCCompile xxx.prototxt -w yyy.caffemodel -o graph -s 6 ”
xxx.prototxt is the structure file of network, yyy.caffemodel is the weight file of
network.
Document “Train_Model.pdf” describes how to modify and generate these two files.
Document “SIMCAM_API.pdf” describes how to read and run graph on borad.
In folder examples/models/, we provide other models for different recognition.
Labelmap.prototxt describes the classifications of the model.

Navigation menu