MNIST Image Classifier from Scratch

This is a custom build project on the famous MNIST dataset , it is a dataset which contains 28 x 28 pixel (black n white) images of hand-drawn numbers .
There are 3 main files here. The Deep_Learning.py file is the most essential one as it contains the Neural Network class, build by me completely from scratch using raw mathamatics and coding , and with Numpy.
The .pkl file essentially contains the weights and biases or you can say the brain of my neural network which is pre trained and got 97.83% Accuracy on testing dataset containing 10000 images.
The Image_classifier.py file is the file where these two things are used togather, the Deep_learning file is imported and the dataset from the other files are imported too then I use it to train my Neural Network which is then shipped to the .pkl file where it is stored and whenever I want to use that I can just comment all the training stuff and just do NN.load_model(file_from_NN_to_be_imported.pkl) .