HandwrittenNotesClassification

A deep learning model to separate Handwritten notes and non notes images with Python and Tensorflow.

Requirements

• Python 3.6+
• TensorFlow 2.0+
• OpenCV 4.0+
• Numpy
• Json

Usage

1. Clone this repository to your local machine.
2. Install the required packages listed in the requirements.txt file by running pip install -r requirements.txt.
3. Place the images you want to classify in the test_images directory.
4. Put your model in the models directory. The model should be saved in the .h5 format.
5. Run the NotesSeparator.py script to classify the images: NotesSeparator.ipynb if you are running in Jupyter Notebook.
6. The classified images will be saved in the output directory.

Example File Structure

	├── classify_notes.py
	├── models
	│   └── notesmodel.h5
	├── output
	│   ├── non_notes
	│   ├── notes
	│   └── results
	│       ├── image1.png
	│       ├── image1.json
	│       ├── image2.png
	│       ├── image2.json
	│       └── results.json
	├── README.md
	└── test_images
	    ├── image1.jpg
	    └── image2.jpg

Model Architecture

The model used for this project is a Convolutional Neural Network (CNN) with the following layers:

1. Conv2D layer with 16 filters and a kernel size of (3,3)
2. MaxPooling2D layer
3. Conv2D layer with 32 filters and a kernel size of (3,3)
4. MaxPooling2D layer
5. Conv2D layer with 16 filters and a kernel size of (3,3)
6. MaxPooling2D layer
7. Flatten layer
8. Dense layer with 256 neurons
9. Dense layer with 1 neuron (output layer) with a sigmoid activation function

Files Saved After Classification

The classified images are saved in either the notes or non_notes directory in the output folder depending on their classification. Additionally, a results.json file is created in the output/results folder that contains the results of the classification for each image.

Contributing

Contributions are welcome. Please fork the repository and submit a pull request.

License

This project is licensed under the MIT License.