Article Per Year (5 Year)
p-Index From 2020 - 2025
0.23
P-Index
This Author published in this journals
All Journal IJID (International Journal on Informatics for Development)
Diniati Ruaika
Unknown Affiliation
Computer Science & IT

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Optimisation of Residual Network Using Data Augmentation and Ensemble Deep Learning for Butterfly Image Classification Diniati Ruaika; Shofwatul Uyun
IJID (International Journal on Informatics for Development) Vol. 12 No. 2 (2023): IJID December
Publisher : Faculty of Science and Technology, UIN Sunan Kalijaga Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14421/ijid.2023.4038

Abstract

Image classification is a fundamental task in vision recognition that aims to understand and categorize an image under a specific label. Image classification needs to produce a quick, economical, and reliable result. Convolutional Neural Networks (CNN) have proven effective for image analysis. However, problems arise due to factors such as the model’s quality, unbalanced training data, overfitting, and layers’ complexity. ResNet50 is a transfer learning-based convolutional neural network model frequently used in many areas, including Lepidopterology. Studies have shown that ResNet50 performs with lower accuracy than other models for classifying butterflies. Therefore, this study aims to optimise the accuracy of ResNet50 using an augmentation approach and ensemble deep learning for butterfly image classification. This study used a public dataset of butterflies from Kaggle. The dataset contains 75 different butterfly species, 9.285 training images, 375 testing images, and 375 validation images. A sequence of transformation functions was applied. The ensemble deep learning was constructed by incorporating ResNet50 with CNN. To measure ResNet50 optimisation, the experimental results of the original dataset and the proposed methods were compared and analysed using evaluation metrics. The research revealed that the proposed method provided better performance with an accuracy of 95%.
Co-Authors Shofwatul ‘Uyun