Features
• XNNPACK support for TensorFlow
™
Lite and ONNX Runtime, with about 20% to 30% performance gain for quantized
networks on a 32-bit system
• TensorFlow
™
Lite 2.11.0 with XNNPACK delegate activated
• ONNX Runtime 1.14.0 with XNNPACK execution engine activated
• OpenCV 4.7.x
• Python
™
3.10.x (enabling Pillow module)
• Coral Edge TPU
™
accelerator native support
–
libedgetpu 2.0.0 (Grouper) aligned with TensorFlow
™
Lite 2.11.0
–
libcoral 2.0.0 (Grouper) aligned with TensorFlow
™
Lite 2.11.0
–
PyCoral 2.0.0 (Grouper) aligned with TensorFlow
™
Lite 2.11.0
•
The X-LINUX-AI OpenSTLinux Expansion Package v5.0.0 is compatible with the Yocto Project
®
build system Mickledore.
It is validated over the
OpenSTLinux Distribution v5.0 on STM32MP157F-DK2 with a USB image sensor, on
STM32MP157F-EV1 with its built-in camera module, and on STM32MP135F-DK with its built-in camera module
• Support for the OpenSTLinux AI package repository allowing the installation of a prebuilt package using apt-* utilities
• Application samples
–
C++ / Python
™
image classification example using TensorFlow
™
Lite based on the MobileNet v1 quantized model
– C++ / Python
™
object detection example using TensorFlow
™
Lite based on the COCO SSD MobileNet v1
quantized model
– C++ / Python
™
image classification example using Coral Edge TPU
™
based on the MobileNet v1 quantized model
and compiled for the
Edge TPU
™
– C++ / Python
™
object detection example using Coral Edge TPU
™
based on the COCO SSD MobileNet v1
quantized model and compiled for the Edge TPU
™
– C++ face recognition application using proprietary model capable of recognizing the face of a known (enrolled)
user. Contact the local STMicroelectronics support for more information about this application or send a request to
edge.ai@st.com
– Python
™
image classification example using ONNX Runtime based on the MobileNet v1 quantized model
– C++ object detection example using ONNX Runtime based on the COCO SSD MobileNet v1 quantized model
– Python
™
object detection example using ONNX Runtime based on the COCO SSD MobileNet v1 quantized model
• Application support for the 720p, 480p, and 272p display configurations
• X-LINUX-AI SDK add-on extending the OpenSTLinux SDK with AI functionality to develop and build an AI application
easily. The X-LINUX-AI SDK add-on supports all the above frameworks. It is available from the X-LINUX-AI product page
Description
X-LINUX-AI is an STM32 MPU OpenSTLinux Expansion Package that targets artificial intelligence for STM32MP1 series
microprocessors. It contains Linux
®
AI frameworks, as well as application examples to get started with some basic use cases
such as computer vision (CV).
The examples provided in X-LINUX-AI use TensorFlow
™
Lite models for image classification based on MobileNet v1, and for
object detection based on the COCO SSD MobileNet v1 model. The face recognition application provided in X-LINUX-AI as a
prebuilt binary is based on models retrained by STMicroelectronics. Contact the local STMicroelectronics support for more
information about this application.
These examples use either the TensorFlow
™
Lite inference engine supporting Python
™
scripting and C/C++ applications, or the
Google Edge TPU
™
accelerator supporting Python
™
scripting and C/C++applications, or also the ONNX Runtime supporting
Python
™
scripting and C/C++applications.
X-LINUX-AI runs on the STM32MP1 series. It is demonstrated on the STM32MP157F-DK2 with a USB image sensor, on the
STM32MP157F-EV1 with its built-in camera module, and on the STM32MP135F-DK with its built-in camera module.
X-LINUX-AI
DB4255 - Rev 6
page 2/6