The problem Digirec solves

The model takes input of handwritten images in the MNIST ("Modified National Institute of Standards and Technology") dataset and produces a prediction of which digit the image represents.

Challenges we ran into

The first major challenge encountered during the building of the model was the lack of diverse training dataset which was solved by data augmentation using which the dataset was increased artificially by modifying the copies of the existing data. The second major challenge faced was the problem of data overfitting. Due to this, the model wasn't able to generalise and predict new data with accuracy therefore, initially our model had a very low and constant val_accuracy which was solved by early stopping due to which we could set high number of epochs.
Dealing with the AI model was also a very novel experience since we were new to machine learning and were most of the time exploring the basics of it.
We also had difficulty achieving an accuracy of 98+ to make the model more efficient.
But overall it was a great experience and we learnt a lot about AI and machine learning during the making of this project.