ShettySach / Kannada-Handwriting-Calculator

A calculator that uses handwritten Kannada digits and operators to calculate the result, using contour detection and CNN model predictions. Made using PyTorch, OpenCV, PIL and CustomTkinter.

Kannada-Handwriting-Calculator

A calculator that uses handwritten ಕನ್ನಡ (Kannada) digits and operators to calculate the result, using contour detection and CNN model prediction.

• Tensorflow (Keras) is used to create, train and load the neural network model used for predictions.
• CustomTKinter is used to provide the GUI.
• OpenCV and Pillow (PIL) are used to read input from the GUI canvas and to obtain contours for individual digits/operators.
• The individual digits/operators are detected and predicted. The predictions are combined into a string and evaluated to get the result.

Contour boxes (green), predicted values (blue) and accuracies (red)

Requirements -

pip install -r requirements.txt

• Tensorflow (Keras)
• OpenCV
• Pillow (PIL)
• Pandas
• Numpy
• CustomTkinter

Instructions -

• Clone the repo and run the Jupyter notebook, MAIN.ipynb or run MAIN.py
• You can use Kannada digits ೦ ೧ ೨ ೩ ೪ ೫ ೬ ೭ ೮ ೯, operators + - × /, decimal point . and parentheses ()
• You can also use ×× for exponentiation and // for floor division

Data

• Kannada MNIST dataset
• Symbol dateset by Irfan Chahyadi