Digit Recognizer

Convolutional Neural Network created using Keras with tensorflow backend. This neural network is used to predict which digit is on a photo.

Training data

Google's SVHN was used to train this model.

Input

Input to the model is a 32x32 image conataining one digit.

Model

Model uses Conv2D layers, along with Dropuot, MaxPooling2D, Flatten and dense layers to categorize photos into categories each representing one digit.

Output

Models output is a digit from 0 to 9 which represents the digit recognized on the photo

Model evaluation

Accuracy	94%
Recall	93%
Precision	93%
F1 Score	94%