The goal of this project is to implement a machine learning model which takes as input an image of a Bengali handwritten digit and outputs the corresponding digit number from the 10 output classes.
Also included an interactive notebook to increase the number of images in the original dataset through data augmentation by applying different transformations.
The project requires latest Python version and the following libraries installed:
- numpy
- pandas
- scikit-learn
- matplotlib
- transformers
- torch
- glob
- os
- sys
- random
- cv2
- tqdm
The dataset used for this project can be accesssed from here.
The dataset is structured into 10 folders, each folder contains images of Bengali digits for the corresponding number. For example, 0 folder contains scanned images of Bengali digit ০.
- 0 --> ০
- 1 --> ১
- 2 --> ২
- 3 --> ৩
- 4 --> ৪
- 5 --> ৫
- 6 --> ৬
- 7 --> ৭
- 8 --> ৮
- 9 --> ৯
The provided Jupyter notebook contains code covering exploratory data analysis, modeling, training, and evaluating model performance.
All codes were run on Kaggle kernel with a GPU.
To open the .ipynb files in your browser and look at the output of the completed cells, use the following command in your terminal after changing the working directory to the project directory bengali-handwritten-digit-classification:
jupyter notebook <file_name>.ipynb