Downloads: 102
Research Paper | Neural Networks | Volume 9 Issue 7, July 2020 | Pages: 346 - 348 | India
Convolution Neural Network: Implementation of a Handwritten Digit Recognition System
Abstract: Digit recognition is one of a famous problem in today's era of different fields like deep learning, machine learning, and computer vision applications. Different techniques are implemented to solve the problem of handwritten digit recognition but this paper focuses on the approach of the neural networks and specifically about the convolution neural network (CNN) approach. The CNN model in this paper is evaluated on many different factors like loss during validation, the accuracy obtained during validation and training. The model trained has a Training Loss = 0.0195 and Validation Loss = 0.0235 due to the similar shapes of some digits like (3, 5), (1, 7), (8, 5), (3, 8) and (6, 9). And the Training Accuracy and Validation Accuracy are calculated as 99.37 and 99.22 respectively.
Keywords: Handwritten digit recognition, Convolution Neural Network, TensorFLow, Image Pre-processing, MNIST Dataset
How to Cite?: Akshit Gambhir, "Convolution Neural Network: Implementation of a Handwritten Digit Recognition System", Volume 9 Issue 7, July 2020, International Journal of Science and Research (IJSR), Pages: 346-348, https://www.ijsr.net/getabstract.php?paperid=SR20602154548, DOI: https://dx.doi.org/10.21275/SR20602154548