loading page

    Impact of Large Datasets on Classification Performance in Deep Learning   
  • +1
  • Davi Nascimento de Paula,
  • Taha Jirjees Ahmed,
  • Navya Basavaraju,
  • Lara Christmann
Davi Nascimento de Paula
Technische Universität Berlin
Author Profile
Taha Jirjees Ahmed
Technische Universität Berlin
Author Profile
Navya Basavaraju
Technische Universität Berlin
Author Profile
Lara Christmann
Technische Universität Berlin
Author Profile

Abstract

Although there are many datasets publicly available for specific domains, there are many possible applications of deep learning where no or limited data is available. Within this project, we analyzed the relationship between the amount of training-data and the accuracy of image classifiers. We used three different datasets with traffic signs images were used for training recent deep learning with different subsets sizes of the original datasets.