Inhaltsverzeichnis

Google Coral

Aneleitung zum Installierern: https://coral.ai/docs/accelerator/get-started/#pycoral-on-linux

Punkt 2a) gibt bei mir (Ubuntu 22.x) einen Fehler bzgl. Abhängigkeiten. Python 3.10 zu NEU!

Lösung:

Quelle1)

sudo apt-get install python3-pycoral

Here is where you should get errors:

The following packages have unmet dependencies:
 python3-pycoral : Depends: python3-tflite-runtime (= 2.5.0.post1) but it is not going to be installed
                   Depends: python3 (< 3.10) but 3.10.6-1~22.04 is to be installed

Instead using apt-get install download 2 wheel files from this repository:

https://github.com/hjonnala/snippets/tree/main/wheels/python3.10

and run:

pip install tflite_runtime-2.5.0.post1-cp310-cp310-linux_x86_64.whl
pip install pycoral-2.0.0-cp310-cp310-linux_x86_64.whl

Coral USB

Object detection and image classification with Google Coral USB Accelerator 2)

Video : https://coral.ai/docs/dev-board/camera/#view-with-a-streaming-server

Use Coral Edge TPUs to run TFlite models in Node with TensorFlow.js: https://codelabs.developers.google.com/tensorflowjs-coral-tflite-node#0

X-LINUX-AI - object detection using Coral Edge TPU TensorFlow Lite C++ API : https://wiki.stmicroelectronics.cn/stm32mpu/wiki/X-LINUX-AI_-_object_detection_using_Coral_Edge_TPU_TensorFlow_Lite_C%2B%2B_API

Beispiele

semantic_segmentation.py

Fehler in 'semantic_segmentation.py':

Zeile 131:

output_img.paste(mask_img, (width, 0))

muss heissen:

output_img.paste(mask_img, (new_width, 0))

Imagemagick - Trimming with a Specific Color: https://legacy.imagemagick.org/Usage/crop/#trim_color

STM32MP1

https://wiki.st.com/stm32mpu/wiki/How_to_compile_model_and_run_inference_on_Coral_Edge_TPU_using_STM32MP1

Models

https://helloworld.co.in/article/model-garden-testing-20-machine-learning-models-raspberry-pi

Projekte

Body segmentation https://blog.tensorflow.org/2019/11/updated-bodypix-2.html