Skip to content

Latest commit

 

History

History
12 lines (10 loc) · 726 Bytes

README.md

File metadata and controls

12 lines (10 loc) · 726 Bytes
Raw

MNIST-Classification

A CNN image recognition model (digit-images)

Description

  • This is a CNN model to which the user feeds an image.
  • The image contains a digit, the model processes this image and outputs the digit in that image.
  • This model is trained using the MNIST dataset.
  • This model works only for images containing single digits, i.e. cannot feed an image containing two digits [1,2], should only have [1] or [2].
  • The model has a LeNet architecture with minor changes to increase training and validation accuracy.

Motivation

The motivation was to learn how to construct a multi-layered CNN and how to train the same, and hence used MNIST because it is a relatively small and easy to work on dataset.