This project shows an example of how to use a Keras model to implement classification using a simple CNN for images.

1. Pull the code
2. Create a Python environment and activate it

e.g. virtualenv -p python3 .venv source .venv/bin/activate

3. Install requirements

e.g. pip install -r requirements.txt

4. There are 2 notebooks (main.ipynb, augment_button/augment_button.ipynb), one for the training process and the other one for the image augmentation process

• The augmentation is done by using the Augmentor package

5. To generate the images, run the generator notebook and copy the files from augment_button/images into dataset/

Value/class names are taken from the folder name in train, which maps the values of the variable MAP_CLASSES in main.ipynb e.g.

6. main.ipynb expects images to be .jpg instead of .png resulting from the augmentation. You can use the following command:

e.g. mogrify -format jpg *.png

7. Once images are augmented, you can manually create a train/test split as you want

• The dataset/train folder has one folder for each class
• The dataset/test folder has all the images dropped-in. Take into account that the class names for this images are taken from the image name
  • e.g. off_1.jpg, on_153.jpg

The classification predicts the state of the button e.g.