HandwrittenDigitClassification

1. Under the course NU-302, 'Research & Development Project', researched and worked on the project for recognition of handwritten digits.

2. Implemented Convolutional Neural Networks (CNN) Deep Learning Model over the MNIST Dataset.

3. A unique feature that was implemented was, that apart from recognizing the normal MNIST dataset black and white images we trained our model to recognize even coloured images and images of any type whether it is written on a ruled sheet, normal page or even it is in a form of pdf also.

4. Another unique feature that was implemented was that we trained our model to handle image data augmentation (that is handle rotated images as well) keeping in mind all the use cases.

5. Achieved 99.62% Training Accuracy and 98.76% Test Accuracy.

6. Our this Research Paper: "Handwritten Digit Classification" has been accepted and published in the International Journal of Emerging Technologies and Innovative Research (JETIR).

7. Published Research Paper Link: https://www.jetir.org/view?paper=JETIR2107649

8. Check out the model performance:

demo.mp4