A CNN image recognition model (digit-images)
- 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.
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.