This is my first project on deep learning. I've used keras and tensorflow to solve the mnist dataset. I've solved this in Google Colab. The link to Colab repository is also given in the python file. The descriptions of the working of the code will be updated soon. An extra function named "evaluate_model()" is commented out. For anyone who wants to check the accuracy of the model using k-fold corss validation feature of scikit learn, just uncomment the code and check it for any number of kfolds you want. If you want to test the model on any image of a digit, just change the path name of the image in "run_example()" function and see how the model predicts the digit in your image. Thank you