ONNX Python image classification

Revision as of 11:02, 22 December 2022 by Registered User

Applicable for STM32MP13x lines, STM32MP15x lines

This article explains how to experiment with ONNX Runtime [1] applications for image classification based on the MobileNet v1 model using ONNX Python™ runtime.

Info white.png Information
Python applications are adequate for prototyping but they are less efficient than C/C++ applications.

1 Description[edit]

The image classification neural network model allows the identification of the subject represented by an image. It classifies an image into various classes.

Python ONNX Runtime image classification application


The application enables three main features:

  • A camera streaming preview implemented using Gstreamer
  • An NN inference based on the camera inputs (or test data pictures) run by the ONNX Runtime [1] interpreter
  • A user interface implemented using Python GTK

The performance depends on the number of CPUs available. The camera preview is limited to one CPU core while the ONNX runtime[1] interpreter is configured to use the maximum of the available resources.

The model used with this application is the MobileNet v1 downloaded from the Tensorflow Lite Hub[2] and converted to ONNX opset 16 format using tf2onnx.

1.1 Convert a Tensorflow™ Lite model to ONNX using tf2onnx[edit]

To convert a .tflite model to the ONNX format, ONNX provides a tool named tf2onnx [3], which is very simple to use.

The first step is to install Tensorflow on the host computer. For test purposes it might be useful to install also ONNX Runtime. The tool tf2onnx uses the versions of Tensorflow and ONNX Runtime already installed. If it does not find any, it installs the most recent versions.

The second step is to install tf2onnx:

- Install from pypi :

 pip install -U tf2onnx

or

- Install latest from GitHub:

 pip install git+https://github.com/onnx/tensorflow-onnx

After the installation, the user is able to convert the .tflite model directly using the following command line:

  python -m tf2onnx.convert --opset 16 --tflite path/to/tflite/model.tflite --output path/to/onnx/model/model.onnx

Native ONNX models are also available in the ONNX Model Zoo [4].

2 Installation[edit]

2.1 Install from the OpenSTLinux AI package repository[edit]

Warning white.png Warning
The software package is provided AS IS, and by downloading it, you agree to be bound to the terms of the software license agreement (SLA0048). The detailed content licenses can be found here.

After having configured the AI OpenSTLinux package, the user can install the X-LINUX-AI components for this application:

 apt-get install onnx-cv-apps-image-classification-python

Then, the user can restart the demo launcher:

- For OpenSTLinux distribution with a version lower than 4.0 use:

 systemctl restart weston@root

- For other OpenSTLinux distribution use:

 systemctl restart weston-launch

2.2 Source code location[edit]

The label_onnx.py Python script is available:

  • in the Openembedded OpenSTLinux Distribution with the X-LINUX-AI Expansion Package:
<Distribution Package installation directory>/layers/meta-st/meta-st-stm32mpu-ai/recipes-samples/onnxrt-cv-apps/files/image-classification/python/label_onnx.py
  • on the target:
/usr/local/demo-ai/computer-vision/onnx-image-classification/python/label_onnx.py
  • on GitHub:
https://github.com/STMicroelectronics/meta-st-stm32mpu-ai/tree/v2.2.0/recipes-samples/onnxrt-cv-apps/files/image-classification/python/label_onnx.py

3 How to use the application[edit]

3.1 Launching via the demo launcher[edit]

Demo launcher

3.2 Executing with the command line[edit]

The Python script label_onnx.py application is located in the userfs partition:

/usr/local/demo-ai/computer-vision/onnx-image-classification/python/label_onnx.py

It accepts the following input parameters:

usage: label_onnx.py [-h] [-i IMAGE] [-v VIDEO_DEVICE] [--frame_width FRAME_WIDTH] [--frame_height FRAME_HEIGHT] [--framerate FRAMERATE]
                     [-m MODEL_FILE] [-l LABEL_FILE] [--input_mean INPUT_MEAN] [--input_std INPUT_STD] [--validation]
                     [--num_threads NUM_THREADS]

options:
  -h, --help            show this help message and exit
  -i IMAGE, --image IMAGE
                        image directory with image to be classified
  -v VIDEO_DEVICE, --video_device VIDEO_DEVICE
                        video device (default /dev/video0)
  --frame_width FRAME_WIDTH
                        width of the camera frame (default is 640)
  --frame_height FRAME_HEIGHT
                        height of the camera frame (default is 480)
  --framerate FRAMERATE
                        framerate of the camera (default is 15fps)
  -m MODEL_FILE, --model_file MODEL_FILE
                        .onnx model to be executed
  -l LABEL_FILE, --label_file LABEL_FILE
                        name of file containing labels
  --input_mean INPUT_MEAN
                        input mean
  --input_std INPUT_STD
                        input standard deviation
  --validation          enable the validation mode
  --num_threads NUM_THREADS
                        Select the number of threads used by ONNX interpreter to run inference

3.3 Testing with MobileNet V1[edit]

The model used for test is the mobilenet_v1_0.5_128_quant.onnx downloaded from Tensorflow Lite Hub[2] and converted to ONNX format.

Info white.png Information
The different objects that the neural network is able to classify are listed in the labels_onnx.txt file located in the target:

/usr/local/demo-ai/computer-vision/models/mobilenet/labels_onnx.txt

To launch the Python script more easily, two shell scripts are available:

  • launch image classification based on camera frame inputs
 /usr/local/demo-ai/computer-vision/onnx-image-classification/python/launch_python_label_onnx_mobilenet.sh
  • launch image classification based on the pictures located in /usr/local/demo-ai/computer-vision/models/mobilenet/testdata directory:
 /usr/local/demo-ai/computer-vision/onnx-image-classification/python/launch_python_label_onnx_mobilenet_testdata.sh
Info white.png Information
The users must populate the testdata directory with their own data sets.

The pictures are then randomly read from the testdata directory

4 References[edit]