Urdu and MNIST Datasets Digit Recognition using CNN Encoder with K-means Clustering

In computer vision and machine learning, hand-written digit recognition is a fundamental issue, which finds potential applications ranging from digital document processing to postal automation. In this paper, a unique convolutional neural network (CNN) encoder followed by K-means clustering strategy is proposed for effective recognition of handwritten digits. Performance of the proposed approach is demonstrated on Urdu handwritten digits dataset and MNIST (Modified National Institute of Standards and Technology database) dataset. The performance is compared with the digits recognition with Principal Component Analysis (PCA) followed by K-means; Auto-encoder followed by K-means. Simulation results reveal proposed approach has superior performance with accuracy, precision and F1-score over the two comparative models.