ailia SDK Tutorial (Python)
Here is a tutorial on how to use ailia SDK in Python. By using ailia SDK and ailia MODELS, you can easily run various AI models.
Installing Python and GPU drivers
macOS
For macOS 12.6 Monterey or later, Python3 is installed as default.
For macOS 12.5 or earlier, Python2 is installed, so use brew to install Python3.
After installing Brew, install Python3 with the following command
brew install python3Windows
For Windows, install Python 3.7 from the following site
Open a command prompt and make sure you can type python3 to run it. If you cannot find the command prompt, type command from the Windows 10 search bar.
In this state, GPU inference using Vulkan is possible.
On nVidia GPUs, you can also use cuDNN. cuDNN requires CUDA Toolkit 9.1 or higher and cuDNN 7.5.0 or higher to be installed. cuDNN requires registration with nVidia Developer.
Linux
On Ubuntu 18.04 LTS, Python3 is already installed, but the package management tool pip is not installed, so we install it.
apt install python3-pipIf you want to use Vulkan, install the Vulkan libraries.
apt install libvulkan1The CPU and Vulkan will work in this state, but if you want to use cuDNN, you need to install CUDA Toolkit and cuDNN.
After downloading cuDNN, you can install it with the following command
sudo dpkg -i libcudnn7_7.6.5.32–1+cuda10.1_amd64.debJetson
Jetson comes with Jetpack and various libraries including cuDNN and OpenCV installed by default. ailia SDK 1.2.3 uses Jetpack4.2 and cuDNN7. ailia SDK 1.2.4 or later also uses Jetpack4.4 and cuDNN8. cuDNN8 are also available. Ubuntu version 18.04LTS or later is supported.
Python3 is already installed, but pip, the package management tool, is not installed, so we will install it.
apt install python3-pipJetpack includes opencv by default, but if you don’t have Jetpack installed and there are errors with opencv, you can install Jetpack by using the following command.
apt install nvidia-jetpackRaspberry Pi
Raspberry Pi OS (32-bit) with desktop and recommended software (
August 2020) download and write using balenaEtcher.
Raspberry Pi OS August 2020 comes with Python 3.7.3 installed.
Install ailia SDK
Installation of the trial version of the ailia SDK
You can use the ailia SDK from Python with just the following command.
pip3 install ailiaThe necessary license file will be automatically downloaded at runtime.
If you need more detailed documentation, you can apply for the trial version of the ailia SDK from the website of ax corporation. When you fill out the application form, the download URL and evaluation license file will be automatically sent to you by email.
Installation of the ailia SDK production version
If you download and extract the ailia SDK, you will see the following layout.
By copying the libraries in the python and library folders included in the SDK, you can use the ailia SDK from Python by following the steps below.
Navigate to the ailia SDK python folder and install it with the pip command.
cd ailia_sdk_version/python
python3 bootstrap.py
pip3 install .Run samples
We use ailia-models for the sample. ailia-models is supported since Python 3.6.
To use ailia MODELS, clone the following site.
git clone https://github.com/axinc-ai/ailia-modelsOf course, you can also download in zip from Clone or Download.
The sample programs use OpenCV, numpy, request, urllib, matplotlib, and scipy.
For non-Jetson users (Windows, Mac, Linux), install it with the following command
pip3 install -r requirements.txtFor Jetson and Raspberry Pi, the multiple libraries contained in requirements.txt are not available, so install them with the following command
Jetson
pip3 install cython
pip3 install numpy
sudo apt-get install python3-matplotlib
sudo apt-get install python3-scipyRaspberry Pi
pip3 install numpy
pip3 install opencv-python
pip3 install matplotlib
pip3 install scikit-image
sudo apt-get install libatlas-base-devNow you can use the ailia SDK from Python.
Navigate to the ailia-models/image_classification/resnet50 folder.
Open a command prompt or terminal in this folder and execute the following command
python3 resnet50.pyYou can then use the ailia SDK to get the inference results.
Each folder in ailia-models contains different types of models, and since there are more than 300 models available, you are encouraged to experiment with them.
Related Information
Please refer to the following tutorial if you want to use the ailia SDK from your browser.
ax Inc. has developed the ailia SDK, which enables cross-platform, GPU-based rapid inference. ax Inc. provides a wide range of services from consulting, model creation, SDK provision of SDKs, development of AI-based applications and systems, to support Please feel free to contact us as we offer a total solution for.
