Handwritten digit recognition using deep learning

Handwritten digit recognition is a great challenge in computer vision, which has application in postal and banking, as human have different type of handwriting, it is difficult to recognize, so precise recognition is required. In this paper, a CNN based approach is used as it effectively recognize handwritten digits as presented. SVM and k-NN methods are traditional used in nature and involve heavy feature engineering and are not capable of handling varying handwriting styles. CNNs, instead, learn automatically hierarchical features from unprocessed images and achieve much higher accuracy. The model includes convolutional layers, pooling layers, full connected layers, and dropout layers to avoid overfitting. The use of optimizer and categorical cross-entropy loss function, the model attains a test accuracy of over 0.98 on the MNIST test set, surpassing the traditional classifiers.