Character recognition using LeNet-5

A deep model (LeNet-5) trained on the MNIST dataset is used for character recognition.

The LeNet-5 model implemented in this project has 3 convolutional layers and 2 fully-connected layers. It has 62,000 training parameters, and the image input size is 32*32. This model achieved 98.48% accuracy on the MNIST test set after training on its train set. MNIST is a dataset of handwritten digits with 70,000 centred fixed-size grey-scale images. More details about the dataset are available in:

http://yann.lecun.com/exdb/mnist

Run the GUI and select your image.