Overfitting Problem in Images Classification for Egg Incubator Using Convolutional Neural Network

Chickens can be hatched using an incubator in addition to hatching naturally through direct hatching by the hen. The incubator process necessitates regular monitoring, including the recording of temperature and humidity, as well as monitoring the images on the incubator, such as the position of the...

Full description

Saved in:
Bibliographic Details
Published in2021 9th International Conference on Cyber and IT Service Management (CITSM) pp. 1 - 7
Main Authors Junaidi, Apri, Ferani Tanjung, Nia Annisa, Wijayanto, Sena, Lasama, Jerry, Iskandar, Ade Rahmat
Format Conference Proceeding
LanguageEnglish
Published IEEE 22.09.2021
Subjects
CNN architecture
Convolution
Convolutional neural networks
dropout regularization
hatching machine
Humidity
Image classification
Monitoring
overfitting
Transfer learning
Online AccessGet full text

Cover

More Information
Summary:Chickens can be hatched using an incubator in addition to hatching naturally through direct hatching by the hen. The incubator process necessitates regular monitoring, including the recording of temperature and humidity, as well as monitoring the images on the incubator, such as the position of the eggs and whether or not chicks have hatched, so that all images on the incubator can be classified. Image classification research for this hatching machine object is divided into three categories: chicks, eggs, and hatched eggs. This study has issues with overfitting the model. To address the issues, several methods were used, including image augmentation, image reshaping, and Dropout regularization on the convolution layer. The researchers made several models with different CNN architectures to see if they could recognize improvements from overfitting. This research obtained an accuracy of 0.8687 after making several models with different CNN architectures.
DOI:10.1109/CITSM52892.2021.9588815