/MNIST

A deep Learning project to classify a handwritten digit based on the dataset from MNIST

Primary LanguageJupyter Notebook

MNIST - Digit Classification

A deep Learning project to classify a handwritten digit based on the dataset from MNIST.

Libraries/Requirements:

  • Python 3.6.9
  • Tensorflow 2.0
  • Keras 2.2.4
  • Numpy 1.15.4
  • Matplotlib 3.0.2

Model architecture:

  1. ANN:
    • Input layer of dimension (28*28)
    • Dense layer of size (28*28) with relu activation
    • Dense layer of size (28*28) with relu activation
    • Dense layer of size 10 with softmax activation for prediction
  2. CNN:
    • Input layer of dimension (28*28)
    • Reshape (28*28) tensor to (28,28,1)
    • Convolution layer of 32 filters with relu activation
    • Maxpooling layer
    • Convolution layer of 64 filters with relu activation
    • Maxpooling layer
    • Convolution layer of 128 filters with relu activation
    • Maxpooling layer
    • Flatten layer
    • Dense layer of size 1024 with relu activation
    • Dense layer of size 10 with softmax activation for prediction

Training and Testing:

The Jupyter Notebook has the related information and code