Recognition of Handwritten Digit Using Neural Networks
Kancherla Santoshi, Chunduru Anilkumar, Suvvari Sravanthirani Bhardwaj · 2022
Because of its practical uses in our daily lives, manuscript number identification is gradually relevant in the present times. Numerous identification techniques such as calligraphy number identification are deployed in many areas where great classification efficiency is required in recent years. In addition pattern recognition is a key component of any computer vision or artificial intelligence system. Machine Learning as well as Computer Vision researchers have used it extensively to build practical applications such as reading computerized bank draft numbers. In this project, we utilized an ANN to develop several surfaces attached development of a thinking computer system with two invisible surfaces to recognize manuscript numbers. The backpropagation method searches for the weight value that delivers the lowest overall error in the network while data processing between layers. Invisible layer neurons employ the ReLU activation function, whereas output neurons use SoftMax. The testing was done using the MNIST handwritten database, which is freely available. Finally retrieved a complete 60,000 digit pictures from the MNIST database for training, validation and testing.