Ethiopian Logo & Trademark Classification And Detection Using Convolutional Neural Network

Abstract

In this paper we propose a method for logo Classification and detection using deep learning. Our recognition pipeline is composed of a logo region proposal followed by a Convolutional Neural Network (CNN) specifically trained for logo classification, even if they are not precisely localized. Experiments are carried out on the Fifteen logo and trademark, and we evaluate the effect on classification performance of synthetic versus real data augmentation, and image pre-processing. Moreover, we systematically investigate the benefits of different training choices such as class balancing, sample-weighting and explicit modeling the background class (i.e. no-logo regions). Experimental results confirm the feasibility of the proposed method, that outperforms the methods in the state of the art. The features extracted are fit into the neural network with 100 epochs, 80/20 splitting ratio, and 0.0001 learning rate. EthioLogoNet model achieved Training accuracy 98.14%, and validation accuracy 98.27. The overall accuracy obtained 98.1% can be detect the logo and trademark detect and classify the types of logo and trademark with a high rate of accuracy.