DEEP LEARNING FOR APPLE AND PEAR RECOGNITION

Authors

  • Vitālijs Žukovs Rezekne Academy of Technologies
  • Ilmārs Apeināns Rezekne Academy of Technologies
  • Sergejs Kodors Rezekne Academy of Technologies

DOI:

https://doi.org/10.17770/het2020.24.6761

Keywords:

AlexNet, apple, CNN, Fruits360, Food2030, neural network, pear

Abstract

The aim of this work is to develop a neural network, which is able to recognize apples and pears. To achieve the goal, the authors of this work used the architecture of the neural network AlexNet and the open dataset “Fruits360”. A trained model showed a good result testing it on validation images: total accuracy 0.97 and latency 35ms/step. In the future research, authors consider to train the neural network model using the MobileNet architecture and verify it using the Cohen`s Kappa coefficient.

Downloads

Download data is not yet available.

References

Bioekonomikas politika [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://ec.europa.eu/research/bioeconomy/index.cfm;jsessionid=1704F40ED0043811F4D22A4F02D1EF7D.cfusion10101?pg=policy&lib=food2030&CFID=23513089&CFTOKEN=a85794a01c992bd5-91D8880F-A7BE-42A6-2E4773E9A2A41389

Konvolūcijas neironu tīkls [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://towardsdatascience.com/a-comprehensive-guide-to-convolutional-neural-networks-the-eli5-way-3bd2b1164a53

AlexNet arhitektūra [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf

Fruits360 datu kopa [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://www.kaggle.com/moltean/fruits

TensorFlow [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://www.tensorflow.org/about

Anaconda rīks [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://www.anaconda.com/why-anaconda/

Cohen`s Kappa [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://www.statisticshowto.com/cohens-kappa-statistic/

MobileNet arhitektūra [tiešsaiste], [atsauce uz 02.04.2020]. Pieejams: https://towardsdatascience.com/review-mobilenetv1-depthwise-separable-convolution-light-weight-model-a382df364b69

Published

2020-04-22

Issue

Section

Information Technologies