/mnist-classifier

Mnist handwritten digit classifier from scratch using numpy only and also with CNN-tf. The aim is to make the models of almost equal accuracy.

Primary LanguagePython

Contains 2 approaches to solving the MNIST Hand Written Digit Classifier.

Main file : Tensorflow Keras API approach. basic.py : Neural Network from scratch.

Architecture remains same. Input : 28 * 28 pixel Grayscale Image. Input layer of Neural Network : 784 Neurons (ReLu activation function) Output : 10 neurons (Softmax Function)

Requirements : 1) Tensorflow, Keras 2) Numpy 3) Pandas