Study on Handwritten Recognition Neural Network Model using Fine Tuning and SELU Activation Function
In this paper, we describe how to implement handwritten recognition system using neural network model. Fine
tuning was applied to the neural network model VGG16 to improve performance. As a result, this method achieved an
accuracy of more than 80%. We also compared the performance of the model using selu as the activation function of VGG16
top model without input data normalization and the model using relu as the activation function of VGG16 top model after
input data normalization. As a result, the loss converges more rapidly when the activation function selu is used, and it is
confirmed that the activation function selu is effective for improving the learning speed.
Keywords - Fine tuning, Handwritten recognition, SELU, VGG16