Last edited 11 months ago

TFLite Python object detection

Applicable for STM32MP13x lines, STM32MP15x lines

This article explains how to experiment with TensorFlow Lite[1] applications for object detection based on the COCO SSD MobileNet v1 model using TensorFlow Lite Python runtime.

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

1. Description[edit source]

The object detection[2] neural network model allows identification and localization of a known object within an image.

Python TensorFlow Lite object detection application

The application enables three main features :

  • A camera streaming preview implemented using Gstreamer
  • A NN inference based on the camera (or test data pictures) inputs is being ran by the TensorFlow Lite[1] interpreter
  • A user interface implemented using Python GTK.

The performances depend on the number of CPUs available. The camera preview is limited to one CPU core where the TensorFlow Lite[1] interpreter is configured to use the maximum of the available resources.

The model used with this application is the COCO SSD MobileNet v1 downloaded from the Tensorflow Lite Hub[3].

2. Installation[edit source]

2.1. Install from the OpenSTLinux AI package repository[edit source]

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 you can install the X-LINUX-AI components for this application:

 apt-get install tflite-cv-apps-object-detection-python

Then restart the demo launcher:

 systemctl restart weston-graphical-session.service

2.2. Source code location[edit source]

The Python script is available:

  • in the Openembedded OpenSTLinux Distribution with X-LINUX-AI Expansion Package:
<Distribution Package installation directory>/layers/meta-st/meta-st-x-linux-ai/recipes-samples/tflite-cv-apps/files/object-detection/python/
  • on the target:
  • on GitHub:

3. How to use the application[edit source]

3.1. Launching via the demo launcher[edit source]

Demo launcher

3.2. Executing with the command line[edit source]

The Python script application is located in the userfs partition:


It accepts the following input parameters:

usage: [-h] [-i IMAGE] [-v VIDEO_DEVICE] [--frame_width FRAME_WIDTH] [--frame_height FRAME_HEIGHT] [--framerate FRAMERATE] [-m MODEL_FILE] [-l LABEL_FILE]
                        [-e EXT_DELEGATE] [-p {std,max}] [--edgetpu] [--input_mean INPUT_MEAN] [--input_std INPUT_STD] [--validation] [--num_threads NUM_THREADS]
                        [--maximum_detection MAXIMUM_DETECTION] [--threshold THRESHOLD]

  -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 ex: 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
                        .tflite model to be executed
  -l LABEL_FILE, --label_file LABEL_FILE
                        name of file containing labels
  -e EXT_DELEGATE, --ext_delegate EXT_DELEGATE
                        external_delegate_library path
  -p {std,max}, --perf {std,max}
                        [EdgeTPU ONLY] Select the performance of the Coral EdgeTPU
  --edgetpu             enable Coral EdgeTPU acceleration
  --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 tflite interpreter to run inference
  --maximum_detection MAXIMUM_DETECTION
                        Adjust the maximum number of object detected in a frame accordingly to your NN model (default is 10)
  --threshold THRESHOLD
                        threshold of accuracy above which the boxes are displayed (default 0.60)
Info white.png Information
The --edgetpu and -p options are only available with a Coral Edge TPU[4] specific model. If you want to use the hardware acceleration please refer to the article provided for this purpose : Coral Python object detection

3.3. Testing with COCO ssd MobileNet v1[edit source]

The model used for test is the detect.tflite downloaded from Tensorflow Lite Hub[3]

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


To ease launching of the Python script, two shell scripts are available:

  • launch object detection based on camera frame inputs
  • launch object detection based on the pictures located in /usr/local/demo-ai/computer-vision/models/coco_ssd_mobilenet/testdata directory
Info white.png Information
Note that you need to populate the testdata directory with your own data sets.

The pictures are then randomly read from the testdata directory

4. References[edit source]