Optimization with ADAM and RMSprop in Convolution neural Network (CNN): A Case study for Telugu Handwritten Characters
International Journal of Emerging Trends in Engineering Research · 2020
Handwritten character recognition is the prime and challenging research area in the area of image processing.This recognition can be defined as the ability of a computer to take the input as handwritten input from various primary sources like paper documents, photographs, touch screens and other devices which are both online and offline.Although a large extent of work has been done in many languages to name a few English and Asian languages such as Japanese, Chinese etc. for handwriting recognition and less amount of work was done on Indian languages like Hindi, Tamil, Telugu, and Kannada etc .This paper explains the methodology for telugu handwritten character recognition algorithm using two optimizers namely Adaptive Moment Estimation (ADAM) and Root Mean Square Propagation (RMSprop) implemented in Convolution neural network (CNN) which has high recognition, accuracy and minimum training and classification time.CNN is dynamic tool to overcome the limitations which had come across in the basic machine learning approaches.To construct the proposed model, we have built our own data set for telugu character (Ka) taking its gunintham as different samples of input.It is observed that RMSprop optimizer outperforms with 89% accuracy compared with ADAM optimizer..