Vehicle Recognition

Transfer learning approach for classification and noise reduction on noisy web data

International Journal of Expert Systems with Applications (ESWA)

Elsevier

Sina Moayed Baharlou

Sina M. Baharlou

S.M Baharlou

Sina Baharlou

Sinabaharlou

Sinabaharlou.com

Sina.baharlou@gmail.com

سینا موید بهارلو

سینا بهارلو

M.Sc in Artificial Intelligence and Robotics from Sapienza University of Rome

Deep Learning, Visual Learning, Convolutional Neural Networks, Machine Learning, Transfer Learning

Gathering an unsupervised image dataset of vehicle models from a marketing site.

Studying the behavior and the performance of six well-known deep architectures in the vehicle recognition case using the transfer learning method on the gathered noisy dataset.

Exploiting the Isolation Forest algorithm with CNN features to discard the external noise of the dataset and analyzing the noise reduction results.

Demonstrating that a high recognition accuracy is attainable with the transfer learning method and it can be elevated using the proposed noise reduction framework.

Providing an android application in which the proposed approach is implemented.