The Vitis AI Library is a set of high-level libraries and APIs built for efficient AI inference with Deep-Learning Processor Unit (DPU). It is built based on the Vitis AI Runtime with Unified APIs, and it fully supports XRT 2020.2.
The Vitis AI Library provides an easy-to-use and unified interface by encapsulating many efficient and high-quality neural networks. This simplifies the use of deep-learning neural networks, even for users without knowledge of deep-learning or FPGAs. The Vitis AI Library allows users to focus more on the development of their applications, rather than the underlying hardware.
For edge users, click Quick Start For Edge to get started quickly.
For cloud users, click Quick Start For Alveo to get started quickly.
Vitis_AI_Library
├── apps
│ ├── seg_and_pose_detect
│ └── segs_and_roadline_detect
└── samples
├── 3Dsegmentation
├── classification
├── covid19segmentation
├── facedetect
├── facefeature
├── facelandmark
├── facequality5pt
├── hourglass
├── lanedetect
├── medicaldetection
├── medicalsegcell
├── medicalsegmentation
├── multitask
├── openpose
├── platedetect
├── platenum
├── pointpillars
├── posedetect
├── refinedet
├── reid
├── retinaface
├── segmentation
├── ssd
├── tfssd
├── yolov2
└── yolov3
-
Download the sdk-2020.2.0.0.sh
-
Install the cross-compilation system environment, follow the prompts to install.
Please install it on your local host linux system, not in the docker system.
./sdk-2020.2.0.0.sh
Note that the ~/petalinux_sdk
path is recommended for the installation. Regardless of the path you choose for the installation, make sure the path has read-write permissions.
Here we install it under ~/petalinux_sdk
.
- When the installation is complete, follow the prompts and execute the following command.
source ~/petalinux_sdk/environment-setup-aarch64-xilinx-linux
Note that if you close the current terminal, you need to re-execute the above instructions in the new terminal interface.
- Download the vitis_ai_2020.2-r1.3.0.tar.gz and install it to the petalinux system.
tar -xzvf vitis_ai_2020.2-r1.3.0.tar.gz -C ~/petalinux_sdk/sysroots/aarch64-xilinx-linux
- To compile the library sample in the AI Library, take
facedetect
as an example, execute the following command.
cd ~/Vitis-AI/demo/Vitis-AI-Library/samples/facedetect
bash -x build.sh
To improve the user experience, the Vitis AI Runtime packages, VART samples, Vitis-AI-Library samples and models have been built into the board image. Therefore, user does not need to install Vitis AI Runtime packages and model package on the board separately. However, users can still install the model or Vitis AI Runtime on their own image or on the official image by following these steps.
-
Installing a Board Image.
-
Download the SD card system image files from the following links:
Note: The version of the board image should be 2020.2 or above.
-
Use Etcher software to burn the image file onto the SD card.
-
Insert the SD card with the image into the destination board.
-
Plug in the power and boot the board using the serial port to operate on the system.
-
Set up the IP information of the board using the serial port. You can now operate on the board using SSH.
-
-
(Optional) Running
zynqmp_dpu_optimize.sh
to optimize the board setting.The script runs automatically by default after the board boots up with the official image. But you can also download the
dpu_sw_optimize.tar.gz
from here.#cd ~/dpu_sw_optimize/zynqmp/ #./zynqmp_dpu_optimize.sh
-
(Optional) How to update Vitis AI Model and install it separately.
For each model, there will be a yaml file which is used for describe all the details about the model. In the yaml, you will find the model's download links for different platforms. Please choose the corresponding model and download it.
Click Xilinx AI Model Zoo to view all the models.- Download the model according to the model's yaml file. Take
densebox_320_320
of ZCU102 as an example.
wget https://www.xilinx.com/bin/public/openDownload?filename=densebox_320_320-zcu102_zcu104-r1.3.0.tar.gz -O densebox_320_320-zcu102_zcu104-r1.3.0.tar.gz
- Copy the downloaded file to the board using scp with the following command.
scp densebox_320_320-zcu102_zcu104-r1.3.0.tar.gz root@IP_OF_BOARD:~/
- Log in to the board (using ssh or serial port) and install the model package.
tar -xzvf densebox_320_320-zcu102_zcu104-r1.3.0.tar.gz cp densebox_320_320 /usr/share/vitis_ai_library/models -r
- Download the model according to the model's yaml file. Take
-
(Optional) How to update Vitis AI Runtime and install them separately.
If you want to update the Vitis AI Runtime or install them to your custom board image, follow these steps.
-
Download the Vitis AI Runtime 1.3.0.
-
Untar the runtime packet and copy the following folder to the board using scp.
tar -xzvf vitis-ai-runtime-1.3.0.tar.gz scp -r vitis-ai-runtime-1.3.0/aarch64/centos root@IP_OF_BOARD:~/
- Log in to the board using ssh. You can also use the serial port to login.
- Install the Vitis AI Runtime. Execute the following command.
cd centos bash setup.sh
-
- Download the vitis_ai_library_r1.3.x_images.tar.gz and the vitis_ai_library_r1.3.x_video.tar.gz. Copy them from host to the target using scp with the following command.
[Host]$scp vitis_ai_library_r1.3.x_images.tar.gz root@IP_OF_BOARD:~/
[Host]$scp vitis_ai_library_r1.3.x_video.tar.gz root@IP_OF_BOARD:~/
- Untar the image and video packages on the target.
cd ~
tar -xzvf vitis_ai_library_r1.3.*_images.tar.gz -C Vitis-AI/demo/Vitis-AI-Library
tar -xzvf vitis_ai_library_r1.3.*_video.tar.gz -C Vitis-AI/demo/Vitis-AI-Library
- Enter the directory of example in target board, take
facedetect
as an example.
cd ~/Vitis-AI/demo/Vitis-AI-Library/samples/facedetect
- Run the image test example.
./test_jpeg_facedetect densebox_320_320 sample_facedetect.jpg
- Run the video test example.
./test_video_facedetect densebox_320_320 video_input.webm -t 8
Video_input.mp4: The video file's name for the input. The user needs to prepare the video file by themselves.
-t: <num_of_threads>
Note that, for examples with video input, only webm
and raw
format are supported by default with the official system image.
If you want to support video data in other formats, you need to install the relevant packages on the system.
- To test the program with a USB camera as input, run the following command:
./test_video_facedetect densebox_320_320 0 -t 8
0: The first USB camera device node. If you have multiple USB camera, the value might be 1,2,3 etc.
-t: <num_of_threads>
- To test the performance of model, run the following command:
./test_performance_facedetect densebox_320_320 test_performance_facedetect.list -t 8 -s 60
-t: <num_of_threads>
-s: <num_of_seconds>
-
Follow Setup Alveo Accelerator Card with HBM for DPUCAHX8H/L to set up the Alveo Card.
-
Download the xclbin files from here. Untar it, choose the Alveo card and install it. Take
U50
as an example.
cd /workspace
wget https://www.xilinx.com/bin/public/openDownload?filename=alveo_xclbin-1.3.0.tar.gz -O alveo_xclbin-1.3.0.tar.gz
tar -xzvf alveo_xclbin-1.3.0.tar.gz
cd alveo_xclbin-1.3.0/U50/6E300M
sudo cp dpu.xclbin hbm_address_assignment.txt /usr/lib
For DPUCAHX8L
, take U50lv
as an example.
cd /workspace
wget https://www.xilinx.com/bin/public/openDownload?filename=alveo_xclbin-1.3.0.tar.gz -O alveo_xclbin-1.3.0.tar.gz
tar -xzvf alveo_xclbin-1.3.0.tar.gz
cd alveo_xclbin-1.3.0/U50lv-V3ME/1E250M
sudo cp dpu.xclbin /opt/xilinx/overlaybins/
export XLNX_VART_FIRMWARE=/opt/xilinx/overlaybins/dpu.xclbin
Note that for DPUCAHX8L
, please refer to XBulter Installation to install XBulter
. After the installation of XBulter
, reboot xbulter service.
sudo service xbutler restart
-
Select the model for your platform.
For each model, there will be a yaml file which is used for describe all the details about the model. In the yaml, you will find the model's download links for different platforms. Please choose the corresponding model and download it.
Click Xilinx AI Model Zoo to view all the models.- Download the model according to the model's yaml file. Take
densebox_320_320
of U50 as an example.
wget https://www.xilinx.com/bin/public/openDownload?filename=densebox_320_320-u50-r1.3.0.tar.gz -O densebox_320_320-u50-r1.3.0.tar.gz
For
DPUCAHX8L
, takeU50lv
as an example.wget https://www.xilinx.com/bin/public/openDownload?filename=densebox_320_320-u50-u50lv-u280-v3me-r1.3.0.tar.gz -O densebox_320_320-u50-u50lv-u280-v3me-r1.3.0.tar.gz
- Install the model package.
If the/usr/share/vitis_ai_library/models
folder does not exist, create it first.
sudo mkdir /usr/share/vitis_ai_library/models
Then install the model package.
tar -xzvf densebox_320_320-u50-r1.3.0.tar.gz sudo cp densebox_320_320 /usr/share/vitis_ai_library/models -r
For
DPUCAHX8L
, execute the following commands to install the model package.tar -xzvf densebox_320_320-u50-u50lv-u280-v3me-r1.3.0.tar.gz sudo cp densebox_320_320 /usr/share/vitis_ai_library/models -r
- Download the model according to the model's yaml file. Take
Note that different alveo cards correspond to different model files, which cannot be used alternately.
Suppose you have downloaded Vitis-AI
, entered Vitis-AI
directory, and then started Docker.
Thus, Vitis-AI-Libray
examples are located in the path of /workspace/demo/Vitis-AI-Library/
in the docker system.
/workspace/demo/Vitis-AI-Library/
is the path for the following example.
If you encounter any path errors in running examples, check to see if you follow the steps above.
- Download the vitis_ai_library_r1.3.x_images.tar.gz and vitis_ai_library_r1.3.x_video.tar.gz packages and untar them.
cd /workspace
wget https://www.xilinx.com/bin/public/openDownload?filename=vitis_ai_library_r1.3.0_images.tar.gz -O vitis_ai_library_r1.3.0_images.tar.gz
wget https://www.xilinx.com/bin/public/openDownload?filename=vitis_ai_library_r1.3.0_video.tar.gz -O vitis_ai_library_r1.3.0_video.tar.gz
tar -xzvf vitis_ai_library_r1.3.0_images.tar.gz -C demo/Vitis-AI-Library/
tar -xzvf vitis_ai_library_r1.3.0_video.tar.gz -C demo/Vitis-AI-Library/
- Enter the directory of sample and then compile it. Take
facedetect
as an example.
cd /workspace/demo/Vitis-AI-Library/samples/facedetect
bash -x build.sh
- Run the image test example.
./test_jpeg_facedetect densebox_320_320 sample_facedetect.jpg
- Run the video test example.
./test_video_facedetect densebox_320_320 video_input.mp4 -t 8
Video_input.mp4: The video file's name for input. The user needs to prepare the video file by themselves.
-t: <num_of_threads>
- To test the performance of model, run the following command:
./test_performance_facedetect densebox_320_320 test_performance_facedetect.list -t 8 -s 60
-t: <num_of_threads>
-s: <num_of_seconds>
For setting up the host for U200/U250 refer to README.
For running Vitis AI Library examples for U200/U250 refer to README.
For more information, please refer to vitis-ai-library-user-guide.