The Relative Performance of Deep Learning and Ensemble Learning for Textile Object Classification


Yildirim P., Birant D.

3rd International Conference on Computer Science and Engineering (UBMK), Sarajevo, Bosnia And Herzegovina, 20 - 23 September 2018, pp.22-26 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/ubmk.2018.8566611
  • City: Sarajevo
  • Country: Bosnia And Herzegovina
  • Page Numbers: pp.22-26
  • Keywords: convolutional neural network, deep learning, ensemble learning, object classification
  • Dokuz Eylül University Affiliated: Yes

Abstract

Object classification is the process of assigning one of the finite sets of classes to objects according to object-level features. Machine learning techniques generally provide accurate prediction results for objects classification task. Therefore, the study presented in this paper proposes a novel advanced neural network architecture that contains convolutional, max pooling, and fully connected layers to classify fashion products. This study also compares the proposed convolutional neural network (CNN) with ensemble learning methods (i.e. Bagging, Random Forest and AdaBoost) in terms of classification accuracy. The results show that the proposed CNN model achieves better classification performance than ensemble learning methods.