EmbeddedMachineLearning

Embedded ML with jupyternotebook and TFLite and STM32CubeIDE.AI

How to Run

In Jupyter Notebook Open Anaconda Prompt.

Launch Jupyter Notebook.

Open lab4.ipynb.

Run the notebook cells in order.

Save Validation Data

The notebook will generate:

Data/images/

Data/models/own_cifar10_model.h5

Data/own_cifar10_validation_20image.csv

Data/labels/own_cifar10_labels.txt

Deploy the Model

Open STM32CubeMX.

Select the B-L475E-IOT01A2 board.

Enable CubeAI under software packs.

Load the .h5 model.

Analyze and generate code.

Open the generated project in STM32CubeIDE.

Flash the board.

Test Inference

Install the required Python packages:

bash python -m pip install -U opencv-python protobuf tqdm==4.50.2 Go to the Misc folder.

Run the AI runner:

bash python ui_python_ai_runner.py Refresh NN and camera on the board tool.

Load the model and label file.

Open an image or use the camera to test classification.