The atlas200dk is ARM based, it runs a copy of ubuntu server in it. It is mainly used to accelerate AI tasks, so it is recommended to separate other backend tasks such as administration, database, etc to a separate backend server. Similar to other single board computers (SBC), the atlas200dk boots from a sd card.
The host machine needs Ubuntu1804 because Mind Studio only works on ubuntu at the time of writing this.
The atlas200dk does not have a wifi or bluetooth interface, so only possible connection is through LAN and type-c USB. Main input power is from the 12v barrel jack. An at least 36W power supply is required.
For more hardware specs, refer here.
Here are the official forums for reference.
The Atlas 200 DK provides an external 10/100/1000M Base-T interface, using the RJ45 connector and connecting to a network with a common network cable.
The Atlas 200 DK provides a Type-C USB port that is compatible with USB 3.0 (SuperSpeed), USB 2.0 (HighSpeed), and USB 1.1 (FullSpeed) communication protocols. This port can be used only in device mode and does not support the master mode. It is used to connect to the debugging host for loading and debugging.
The Atlas 200 DK provides a MicroSD card port that supports SD 3.0 and is backward compatible with SD 2.0. You are advised to use a standard MicroSD card of SD 3.0. The minimum capacity is 8 GB and the maximum capacity is 2 TB.
The Atlas 200 DK has two MIPI-CSI interfaces, and the definitions of the two interfaces are the same.
| Pin | Name | Level | Pin | Name | Level |
|---|---|---|---|---|---|
| 1 | +3.3V | 3.3V | 2 | +5V | 5V |
| 3 | I2C_2-SDA | 3.3V | 4 | +5V | 5V |
| 5 | I2C_2-SCL | 3.3V | 6 | GND | - |
| 7 | GPIO0 | 3.3V | 8 | UART_0-TX | 3.3V |
| 9 | GND | - | 10 | UART_0-RX | 3.3V |
| 11 | GPIO1 | 3.3V | 12 | NC | - |
| 13 | NC | - | 14 | GND | - |
| 15 | GPIO2 | 3.3V | 16 | UART_1-TX | 3.3V |
| 17 | +3.3V | 3.3V | 18 | UART_1-RX | 3.3V |
| 19 | SPI-MOSI | 3.3V | 20 | GND | - |
| 21 | SPI-MISO | 3.3V | 22 | NC | - |
| 23 | SPI-CLK | 3.3V | 24 | SPI-CS | 3.3V |
| 25 | GND | - | 26 | GPIO10 | 3.3V |
| 27 | GPIO8 | 3.3V | 28 | GPIO9 | 3.3V |
| 29 | GPIO3 | 3.3V | 30 | GND | - |
| 31 | GPIO4 | 3.3V | 32 | NC | - |
| 33 | GPIO5 | 3.3V | 34 | GND | - |
| 35 | GPIO6 | 3.3V | 36 | +1.8V | 1.8V |
| 37 | GPIO7 | 3.3V | 38 | UART_Hi35559-TXD | 3.3V |
| 39 | GND | - | 40 | UART_Hi35559-RXD | 3.3V |
UART 0 is an 8-pin or 10-pin interface used for the default debugging serial port (console) of the Ascend 310. The baud rate is 115200 bit/s. UART 1 is a 16-pin or 18-pin interface used for expansion and communication with other modules. UART-Hi3559 has 38 TX pins and 40 RX pins, which are used for the MIPI-CSI interface to access the Hi3559 chip for debugging. The baud rate is 115200 bit/s.
This 4-wire SPI interface only support master mode.
Only I2C_2 can be used, I2C_1 is intra-board interface. Maximum rate is 400kHz.
As output pins, GPIO0, GPIO1, and GPIO2 must be connected to external pull- up resistors to increase driving capabilities. It is recommended that the value of pull-up resistance is 1–10 kilohm.
Latest SDK version at the time writing this is 20.1.0. Setup consists of several steps, mainly software and hardware. For hardware part, we have to flash an SD card with Ubuntu Server 18.04.05 for ARM. For software part, we have to setup some SDks and MindStudio, an IDE with tools to program the atlas200dk.
A computer/vm with Ubuntu18.04. SD card (>=32GB recommended). Huawei Atlas200 development kit.
Ubuntu18.04.04 or Ubuntu18.04.05 is required. Install it in a VM (recommended) or dual boot your computer if you do not have it.
Install dependencies
sudo apt-get update && sudo apt-get upgrade
sudo apt-get install -y git gcc g++ make cmake unzip zlib1g zlib1g-dev libsqlite3-dev openssl libssl-dev libffi-dev pciutils net-tools g++-5-aarch64-linux-gnu g++-aarch64-linux-gnu libblas-dev gfortran libblas3 libopenblas-dev libcanberra-gtk-module
Update default python version
wget https://www.python.org/ftp/python/3.7.5/Python-3.7.5.tgz
tar -zxvf Python-3.7.5.tgz
cd Python-3.7.5
./configure --prefix=/usr/local/python3.7.5 --enable-shared
make -j8
sudo make install
# link new python
sudo cp /usr/local/python3.7.5/lib/libpython3.7m.so.1.0 /usr/lib
sudo ln -s /usr/local/python3.7.5/bin/python3 /usr/bin/python3.7
sudo ln -s /usr/local/python3.7.5/bin/pip3 /usr/bin/pip3.7
sudo ln -s /usr/local/python3.7.5/bin/python3 /usr/bin/python3.7.5
sudo ln -s /usr/local/python3.7.5/bin/pip3 /usr/bin/pip3.7.5
# Install dependencies
pip3.7.5 install attrs psutil decorator numpy protobuf==3.11.3 scipy sympy cffi grpcio grpcio-tools requests --userGet required files , it is recommended to use wget to download these files. Use here if links are not working
| File | URL |
|---|---|
| Ascend-cann-toolkit_20.1.rc1_linux-aarch64.run | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/cann/V100R020C10/Ascend-cann-toolkit_20.1.rc1_linux-aarch64.run |
| Ascend-cann-toolkit_20.1.rc1_linux-x86_64.run | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/cann/V100R020C10/Ascend-cann-toolkit_20.1.rc1_linux-x86_64.run |
| A200dk-npu-driver-20.1.0-ubuntu18.04-aarch64-minirc.tar.gz | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/atlas200dk/20.1/A200dk-npu-driver-20.1.0-ubuntu18.04-aarch64-minirc.tar.gz |
| MindStudio_2.0.0-beta2_ubuntu18.04-x86_64.tar.gz | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/mindstudio/2.0.0-beta2/MindStudio_2.0.0-beta2_ubuntu18.04-x86_64.tar.gz |
Give required permissions and execute file
chmod 755 *.run
./Ascend-cann-toolkit_20.1.rc1_linux-aarch64.run --install
./Ascend-cann-toolkit_20.1.rc1_linux-x86_64.run --installMove the file A200dk-npu-driver-20.1.0-ubuntu18.04-aarch64-minirc.tar.gz into $HOME/Ascend directory. Install the media driver.
cd $HOME/Ascend
tar -zxvf A200dk-npu-driver-20.1.0-ubuntu18.04-aarch64-minirc.tar.gz
Install MindStudio
cd $HOME
tar -zxvf MindStudio_2.0.0-beta2_ubuntu18.04-x86_64.tar.gz
cd MindStudio-ubuntu/bin
./MindStudio.shThere might be instructions for you to install some more dependencies, install them with the command provided and rerun the line.
After this, a window will pop out, choose do not import settings.
At toolkit version settings, choose the path toyour toolkit installed, which is $HOME/Ascend/ascend-toolkit/20.1.rc1.
Congratulations, you have successfully set up your environment for development.
Get tools from here.
| File | URL |
|---|---|
| make_sd_card.py | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/atlas200dk/20.1/make_sd_card.py |
| make_ubuntu_sd.sh | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/atlas200dk/20.1/make_ubuntu_sd.sh |
| ubuntu-18.04.5-server-arm64.iso | https://cdimage.ubuntu.com/releases/18.04/release/ubuntu-18.04.5-server-arm64.iso |
| A200dk-npu-driver-20.1.0-ubuntu18.04-aarch64-minirc.tar.gz | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/atlas200dk/20.1/A200dk-npu-driver-20.1.0-ubuntu18.04-aarch64-minirc.tar.gz |
| Ascend-cann-minirc_20.1.rc1_ubuntu18.04-aarch64.zip | https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resource/cann/V100R020C10/Ascend-cann-minirc_20.1.rc1_ubuntu18.04-aarch64.zip |
Download files listed above and put them in a directory. A SD Card reader is required for this process. If you do not have one, follow instructions here to place a jumper on the atlas200dk to make it into a SD Card reader.
pip3 install pyyaml
sudo apt-get install -y qemu-user-static binfmt-support python3-yaml gcc-aarch64-linux-gnu g++-aarch64-linux-gnu # Install dependencies
sudo chmod 755 make_sd_card.py make_ubuntu_sd.sh ubuntu-18.04.5-server-arm64.iso # give required previliges to execute file
sudo python3 make_sd_card.py local /dev/sdb # change /dev/sdb to your sd card location, you can check with fdisk -lIn this process, the script will help you install other dependencies, it might prompt you to answer 'y'.
Wait about 10 minutes, Make SD Card successfully! means you successfully flashed the ubuntu server OS into the SD card. Plug it into the atlas200dk and you are ready to go.
There are two main ways to connect, through the usb type-c and through RJ45 LAN cable.
Connect the board to your pc with a usb type-c cable. If you are using a VM, you may need to manually connect the device (Huawei Ethernet Gadget) to the VM. Once connected, run
ifconfigYou will get an output similar to this
ens33: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 192.168.0.142 netmask 255.255.255.0 broadcast 192.168.0.255
inet6 fe80::2882:37a5:4e8d:c700 prefixlen 64 scopeid 0x20<link>
ether 00:0c:29:dc:2c:b2 txqueuelen 1000 (Ethernet)
RX packets 3966648 bytes 5866216386 (5.8 GB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 1305635 bytes 111797852 (111.7 MB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
ens160u1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet6 fe80::f07d:6458:cc20:c4df prefixlen 64 scopeid 0x20<link>
ether fe:99:ea:f0:aa:22 txqueuelen 1000 (Ethernet)
RX packets 0 bytes 0 (0.0 B)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 12 bytes 2050 (2.0 KB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
lo: flags=73<UP,LOOPBACK,RUNNING> mtu 65536
inet 127.0.0.1 netmask 255.0.0.0
inet6 ::1 prefixlen 128 scopeid 0x10<host>
loop txqueuelen 1000 (Local Loopback)
RX packets 19188 bytes 1750969 (1.7 MB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 19188 bytes 1750969 (1.7 MB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
The first one(ens33) is your internet connection, the last one(lo) is local network, the centre one(ens160u1) is the atlas200dk. As you can see, it haven't got an IP address assigned to it, so we will assign one to it. To do this, we have to modify netplan.
Modify /etc/netplan/01-network-manager-all.yaml as below with your favourite text editor. Note that .yaml files are indent-sensitive.
# Let NetworkManager manage all devices on this system
network:
version: 2
renderer: NetworkManager
ethernets:
ens160u1: # replace with your network adapter name
dhcp4: no # close dchp4
addresses: [192.168.1.223/8] # Setup IP address and mask for device
gateway4: 255.255.255.0 #设置网关
nameservers:
addresses: [114.114.114.114]and then execute
sudo netplan apply
Run ifconfig again and you will see an IP address assigned to the atlas200dk.
ens33: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 192.168.0.142 netmask 255.255.255.0 broadcast 192.168.0.255
inet6 fe80::2882:37a5:4e8d:c700 prefixlen 64 scopeid 0x20<link>
ether 00:0c:29:dc:2c:b2 txqueuelen 1000 (Ethernet)
RX packets 3967650 bytes 5866333120 (5.8 GB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 1306169 bytes 111864657 (111.8 MB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
ens160u1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 192.168.1.223 netmask 255.0.0.0 broadcast 192.255.255.255
inet6 fe80::fc99:eaff:fef0:aa22 prefixlen 64 scopeid 0x20<link>
ether fe:99:ea:f0:aa:22 txqueuelen 1000 (Ethernet)
RX packets 0 bytes 0 (0.0 B)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 19 bytes 3910 (3.9 KB)
TX errors 174 dropped 0 overruns 0 carrier 0 collisions 0
lo: flags=73<UP,LOOPBACK,RUNNING> mtu 65536
inet 127.0.0.1 netmask 255.0.0.0
inet6 ::1 prefixlen 128 scopeid 0x10<host>
loop txqueuelen 1000 (Local Loopback)
RX packets 20731 bytes 1907482 (1.9 MB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 20731 bytes 1907482 (1.9 MB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
Now, you have successfully established a connection to the atlas200dk. Next, we need to ssh into the board to do some initial setup. Default password is Mind@123
Here, we need to add the path of libascendcl.so into /etc/ld.so.conf.d/mind_so.conf.
su root
echo "/home/HwHiAiUser/Ascend/acllib/lib64/libascendcl.so" >> /etc/ld.so.conf.d/mind_so.conf
ldconfig
exitThen, we need to restart the ada service. Run
ps -ef | grep adaand run (replace <process id> with the id of the first process in output of above command.)
kill <process id>Then, restart the process with
/var/ada
Congratulations, you have completed hardware setup. Now, let's try a demo project.
To start MindStudio, run
sh ./MindStudio-ubuntu/bin/MindStudio.shSelect Create New Project or File -> New -> Project if you have opened other projects before.
Select Ascend App on the left menu. Here, enter details of your project or leave it as it is. Click Next.
Select Sample ResNet-50 and click Finish. Lets see what is in the sample project.
MyApp
├── build # Here are the cmake build files.
│ └── cmake
│ ├── CMakeCache.txt
│ ├── CMakeFiles
│ │ ├── 3.10.2
│ │ │ ├── CMakeCCompiler.cmake
│ │ │ ├── CMakeCXXCompiler.cmake
│ │ │ ├── CMakeDetermineCompilerABI_C.bin
│ │ │ ├── CMakeDetermineCompilerABI_CXX.bin
│ │ │ ├── CMakeSystem.cmake
│ │ │ ├── CompilerIdC
│ │ │ │ ├── a.out
│ │ │ │ ├── CMakeCCompilerId.c
│ │ │ │ └── tmp
│ │ │ └── CompilerIdCXX
│ │ │ ├── a.out
│ │ │ ├── CMakeCXXCompilerId.cpp
│ │ │ └── tmp
│ │ ├── cmake.check_cache
│ │ ├── CMakeDirectoryInformation.cmake
│ │ ├── CMakeOutput.log
│ │ ├── CMakeTmp
│ │ ├── feature_tests.bin
│ │ ├── feature_tests.c
│ │ ├── feature_tests.cxx
│ │ ├── Makefile2
│ │ ├── Makefile.cmake
│ │ ├── progress.marks
│ │ └── TargetDirectories.txt
│ ├── cmake_install.cmake
│ ├── compile_commands.json
│ ├── Makefile
│ └── src
│ ├── CMakeFiles
│ │ ├── CMakeDirectoryInformation.cmake
│ │ ├── main.dir
│ │ │ ├── build.make
│ │ │ ├── cmake_clean.cmake
│ │ │ ├── DependInfo.cmake
│ │ │ ├── depend.make
│ │ │ ├── flags.make
│ │ │ ├── link.txt
│ │ │ └── progress.make
│ │ └── progress.marks
│ ├── cmake_install.cmake
│ └── Makefile
├── CMakeLists.txt
├── data # Here is our input data, two images
│ ├── dog1_1024_683.jpg
│ └── dog2_1024_683.jpg
├── inc # Here are the model processing, init and deinit header files.
│ ├── model_process.h
│ ├── sample_process.h
│ └── utils.h
├── MyApp.iml
├── out
├── script
│ └── transferPic.py # This is a script to covert .jpg images to .bin binary files, and perform resize..
└── src
├── acl.json # System init configs
├── CMakeLists.txt
├── main.cpp # Main function file where the classification task (model inference) is defined
├── model_process.cpp # Model processing functions
├── sample_process.cpp # Data init and deinit
└── utils.cpp # common utilsNow, we need to download the model definitions and weights.
mkdir caffe_model && cd caffe_model
wget https://obs-model-ascend.obs.cn-east-2.myhuaweicloud.com/resnet50/resnet50.caffemodel
wget https://obs-model-ascend.obs.cn-east-2.myhuaweicloud.com/resnet50/resnet50.prototxtNow, we need to convert the caffe models to .om format for the atlas200dk. On the top menu bar, select Ascend -> Model Converter and enter the file path. Other parameters will be loaded automatically based on the model file.
Choose FP16 for data type and click Next. Data preprocessing is not required in this demo, so you can disable it be switching the slider. For the advanced options, we will leave it as it is for this demo. Click Finish. The output tab below will print out some logs.
2020-12-21 00:20:17 Model converted successfully.Something similar to this means a successful conversion. Now, we need to add the model into our project.
Right click your project name, MyApp in my case on the left menu and select Add Model. The converted model can be found in $HOME/modelzoo, which is $HOME/modelzoo/resnet50/device/resnet50.om for my case. Select the model file and click Ok. You can see the model file in the model directory in your project directory now.
Next, we need to convert the two input images located in the data directory into .bin format and resize it. The script transferPic.py in the script directory will do this for us.
Open a terminal in the data directory and run
python3.7.5 ../script/transferPic.pyAfter this, you will see two .bin files created in the data directory.
Now, its time to build the project. Go to Build -> Build Configurations in the top menu. Change the Target Architecture to aarch64. Then, click Build.
After building, you will see a main file generated in the out directory.
Next, connect the atlas200dk to MindStudio. Please make sure you have already setup Netplan and can successfully ssh into the board.
In the top menu, select Ascend -> Device Manager. Click the plus button on the top-right side of the popup window. Key in the IP address of the atlas200dk, which is 192.168.1.2 by default and click Ok. Now, you should see your device connected.
Click Ok to exit device manager. Then, select Run -> Run 'MyApp' to execute the project on the atlas200dk. A window will popup to confirm the configurations, click Ok. You can also run by clicking the green play button on the top menu.
You will get an output similar to this
[INFO] acl init success
[INFO] open device 0 success
[INFO] create context success
[INFO] create stream success
[INFO] get run mode success
[INFO] load model ../model/resnet50.om success
[INFO] create model description success
[INFO] create model output success
[INFO] start to process file:../data/dog1_1024_683.bin
[INFO] model execute success
[INFO] top 1: index[161] value[0.767578] #Here are the results
[INFO] top 2: index[162] value[0.154785] #Here are the results
[INFO] top 3: index[167] value[0.038513] #Here are the results
[INFO] top 4: index[163] value[0.021606] #Here are the results
[INFO] top 5: index[166] value[0.011658] #Here are the results
[INFO] output data success
[INFO] start to process file:../data/dog2_1024_683.bin
[INFO] model execute success
[INFO] top 1: index[267] value[0.936035] #Here are the results
[INFO] top 2: index[266] value[0.041138] #Here are the results
[INFO] top 3: index[265] value[0.018845] #Here are the results
[INFO] top 4: index[219] value[0.002609] #Here are the results
[INFO] top 5: index[160] value[0.000295] #Here are the results
[INFO] output data success
[INFO] unload model success, modelId is 1
[INFO] execute sample success
[INFO] end to destroy stream
[INFO] end to destroy context
[INFO] end to reset device is 0
[INFO] end to finalize aclCongratulations, you have completed your first project.
This projects runs ffmpeg and opencv on atlas200dk. To accomplish this, we need to connect the atlas200dk to the internet to download the required packages. Connect your board to your router with a RJ45 cable.
Firstly, ssh into the board. Default password is Mind@123
We need to configure netplan on the board, the only text editor on board is vim. Unfortunately, you can't install other text editors before you connect to the internet. If you are totally new to vim, here are a few tips that are enough for you to edit the file:
ito switch to edit modeesckey to exit edit mode and go back to command mode:wqto save and exit:qto exit.:q!to force exit(without saving changes).- Note that
!is a force command. - Use arrow keys to navigate around.
su root
vi /etc/netplan/01-netcfg.yamlenable dhcp like below
network:
version: 2
# renderer: NetworkManager
renderer: networkd
ethernets:
eth0:
dhcp4: yes
usb0:
dhcp4: no
addresses: [192.168.1.2/24]
gateway4: 192.168.0.1and run
netplan apply
then, modify /etc/apt/sources.list to the ubuntu-arm sources
deb http://ports.ubuntu.com/ bionic main restricted universe multiverse
deb-src http://ports.ubuntu.com/ bionic main restricted universe multiverse
deb http://ports.ubuntu.com/ bionic-updates main restricted universe multiverse
deb-src http://ports.ubuntu.com/ bionic-updates main restricted universe multiverse
deb http://ports.ubuntu.com/ bionic-security main restricted universe multiverse
deb-src http://ports.ubuntu.com/ bionic-security main restricted universe multiverse
deb http://ports.ubuntu.com/ bionic-backports main restricted universe multiverse
deb-src http://ports.ubuntu.com/ bionic-backports main restricted universe multiverse
deb http://ports.ubuntu.com/ubuntu-ports/ bionic main universe restricted
deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic main universe restricted
then, install dependencies
apt-get update
apt-get install build-essential libgtk2.0-dev libavcodec-dev libavformat-dev libjpeg-dev libtiff5-dev libswscale-dev git cmake libswscale-dev python3-setuptools python3-dev python3-pip pkg-config ffmpeg libopencv-dev -y
pip3 install --upgrade pip
pip3 install Cython
pip3 install numpy
exitCreate dierctory to store built .so files, which we will need in mindstudio to program the atlas200dk afterwards. We need to build these on the atlas200dk because it needs to be built for aarch64, not x86 which our pc probably is.
mkdir -p /home/HwHiAiUser/ascend_ddk/armInstall ffmpeg
cd $HOME
git clone https://git.ffmpeg.org/ffmpeg.git ffmpeg
cd ffmpeg
./configure --enable-shared --enable-pic --enable-static --disable-yasm --prefix=/home/HwHiAiUser/ascend_ddk/arm
make -j8
su root
make install
echo "/home/HwHiAiUser/ascend_ddk/arm/lib" >> /etc/ld.so.conf.d/ffmpeg.conf # for opencv to locate ffmpeg
ldconfig
echo "export PATH=$PATH:/home/HwHiAiUser/ascend_ddk/arm/bin" >> /etc/profile # add ffmpeg to PATH
source /etc/profile
cp /home/HwHiAiUser/ascend_ddk/arm/lib/pkgconfig/* /usr/share/pkgconfig # copy package config filesInstall opencv
cd $HOME
git clone https://github.com/opencv/opencv.git
git clone https://github.com/opencv/opencv_contrib.git
cd opencv
mkdir build
cd build
cmake -D BUILD_SHARED_LIBS=ON -D BUILD_opencv_python3=YES -D BUILD_TESTS=OFF -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/home/HwHiAiUser/ascend_ddk/arm -D WITH_LIBV4L=ON -D OPENCV_EXTRA_MODULES=../../opencv_contrib/modules -D PYTHON3_LIBRARIES=/usr/lib/python3.6/config-3.6m-aarch64-linux-gnu/libpython3.6m.so -D PYTHON3_NUMPY_INCLUDE_DIRS=/usr/local/lib/python3.6/dist-packages/numpy/core/include -D OPENCV_SKIP_PYTHON_LOADER=ON ..
make -j8
su root
make install
cp /home/HwHiAiUser/ascend_ddk/arm/lib/python3.6/dist-packages/cv2.cpython-36m-aarch64-linux-gnu.so /usr/lib/python3/dist-packages
echo "export LD_LIBRARY_PATH=/home/HwHiAiUser/Ascend/acllib/lib64:/home/HwHiAiUser/ascend_ddk/arm/" >> ~/.bashrc
source ~/.bashrc
echo "/home/HwHiAiUser/ascend_ddk/arm/lib/libopencv_core.so.4.5
" >> /etc/ld.so.conf.d/opencv.conf
ldconfigAfter this, we need to sync the built libaries to our development environment, MindStudio.
In your dev environment, run
mkdir $HOME/ascend_ddk
scp -r [email protected]:/home/HwHiAiUser/ascend_ddk/arm $HOME/ascend_ddk/ # copy files to local dev environment
cd /usr/lib/aarch64-linux-gnu
sudo scp -r [email protected]:/lib/aarch64-linux-gnu/* ./
sudo scp -r [email protected]:/usr/lib/aarch64-linux-gnu/* ./Now, we download the project files from https://www.github.com/ascend/samples. This project utilises a CNN to turn grayscale images into coloured LAB images, then it is converted into RGB format.
cd $HOME/AscendProjects
wget https://c7xcode.obs.cn-north-4.myhuaweicloud.com/code_Ascend/colorization.zip
unzip colorization.zipGet models
mkdir -p $HOME/AscendProjects/colorization/caffe_model
cd $HOME/AscendProjects/colorization/caffe_model
wget https://obs-model-ascend.obs.cn-east-2.myhuaweicloud.com/colorization/colorization.caffemodel
wget https://obs-model-ascend.obs.cn-east-2.myhuaweicloud.com/colorization/colorization.prototxtOpen MindStudio with
sh $HOME/MindStudio-ubuntu/bin/MindStudio.shSelect File -> Open on the top menu bar. Open the courization project. The first time you import the project, a window will popup to confirm the project settings.
Then, we convert the caffe models downloaded to .om format with the model converter. The process is same as the previous sample. The data type this time is FP32.
After converting model, load the model into your project. This process is also same as previous project.
Due to some bugs, we have to modify the main.cpp. Comment out these lines
// if((argc < 2) || (argv[1] == nullptr)){
// ERROR_LOG("Please input: ./main <image_dir>");
// return FAILED;
// }and change the line
string inputImageDir = string(argv[1]);to
string inputImageDir = "../data";Then, build the project. This is also same as previous project. Remember to set the target architecture to aarch64.
After build success, connect your board with device manager and run it. As we have set up LAN connection just now, we can connect to the board through LAN. Log into your router to find out what IP is assigned to the board. The idendifier of the board is davinci-mini.
In the device manager, you just need to change the ip to your board's ip and connect. Click refresh to see changes reflected on screen.
Then, run the output image can be found in output directory in the out directory where the compiled binary is located.
Get code
wget https://c7xcode.obs.cn-north-4.myhuaweicloud.com/code\_Ascend/facedetection.zip
unzip facedetection.zipthen open it with mindstudio.
Get models
cd facedetection
mkdir caffemodel && cd caffemodel
wget https://c7xcode.obs.cn-north-4.myhuaweicloud.com/models/face\_detection/face\_detection.caffemodel
wget https://c7xcode.obs.cn-north-4.myhuaweicloud.com/models/face\_detection/face\_detection.prototxtin the .workspace file, change the hostip to the ip of your board
<configuration default="true" type="Ascend App Configuration Type" factoryName="Ascend App" hostip="192.168.0.197" commandArguments="" debugPort="">
<method v="2" />
</configuration>
<configuration name="facedetection" type="Ascend App Configuration Type" factoryName="Ascend App" hostip="192.168.0.197" commandArguments="" debugPort="">
<method v="2" />
</configuration>Convert models and add models as previous examples.
In this sample, we will need a presenter server to receive the camera stream. It is a websockets server served in the ubuntu environment.
Install presenter server.
#Download files
cd $HOME
wget https://c7xcode.obs.cn-north-4.myhuaweicloud.com/presenteragent/presenteragent.zip --no-check-certificate
unzip presenteragent.zip
# Download dependencies
sudo apt-get install autoconf automake libtool python3-pip
python3 -m pip install pip --user
python3 -m pip install --upgrade pip --user
python3 -m pip install tornado==5.1.0 protobuf==3.5.1 numpy==1.14.2 --user
python3.7.5 -m pip install tornado==5.1.0 --user
# Install protobuf, note that this version is REQUIRED. We compile two times, one for your dev environment, one for the atlas200.
git clone -b 3.8.x https://github.com/protocolbuffers/protobuf.git
cd protobuf
./autogen.sh
bash configure
make -j8
sudo make install
make distclean
./configure --build=x86_64-linux-gnu --host=aarch64-linux-gnu --with-protoc=protoc
make -j8
sudo make install
sudo ldconfig
# set environment variables
export DDK_PATH=$HOME/Ascend/ascend-toolkit/20.1.rc1/acllib_linux.aarch64
# Install Presenter Agent.
cd $HOME/presenteragent/
make -j8
make install
# Upload the compiled .so file to the Atlas 200 DK (replace the ip with your board ip).
scp $HOME/ascend_ddk/arm/lib/libpresenteragent.so [email protected]:/home/HwHiAiUser
ssh [email protected]
cp /home/HwHiAiUser/libpresenteragent.so /home/HwHiAiUser/ascend_ddk/arm/libChange the values of presenter_server_ip and presenter_view_ip in data/param.conf to the IP address of your dev environment, and change the value of presenter_agent_ip to the IP address of the atlas200dk.
Build as if in the previous examples.
Start a terminal in the project directory, you can use the built in terminal in MindStudio. Then start the presenter server with
bash script/run_presenter_server.shYou will see a line Please visit <url> for face detection. Go to that url in your browser to see the presenter UI.
Run the built file like previous projects.
In the presenter agent UI, click the person under View Name to see the camera stream.
Congratulations, you have completed another sample.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent