Cloud type classification using deep learning with cloud images

Mehmet Guzel, Muruvvet Kalkan, Erkan Bostanci, Koray Acici, Tunc Asuroglu

Research output: Contribution to journalArticleScientificpeer-review

1 Citation (Scopus)

Abstract

Clouds play a pivotal role in determining the weather, impacting the daily lives of everyone. The cloud type can offer insights into whether the weather will be sunny or rainy and even serve as a warning for severe and stormy conditions. Classified into ten distinct classes, clouds provide valuable information about both typical and exceptional weather patterns, whether they are short or long-term in nature. This study aims to anticipate cloud formations and classify them based on their shapes and colors, allowing for preemptive measures against potentially hazardous situations. To address this challenge, a solution is proposed using image processing and deep learning technologies to classify cloud images. Several models, including MobileNet V2, Inception V3, EfficientNetV2L, VGG-16, Xception, ConvNeXtSmall, and ResNet-152 V2, were employed for the classification computations. Among them, Xception yielded the best outcome with an impressive accuracy of 97.66%. By integrating artificial intelligence technologies that can accurately detect and classify cloud types into weather forecasting systems, significant improvements in forecast accuracy can be achieved. This research presents an innovative approach to studying clouds, harnessing the power of image processing and deep learning. The ability to classify clouds based on their visual characteristics opens new avenues for enhanced weather prediction and preparedness, ultimately contributing to the overall accuracy and reliability of weather forecasts.

Original languageEnglish
Article numbere1779
JournalPeerJ Computer Science
Volume10
DOIs
Publication statusPublished - 2024
MoE publication typeA1 Journal article-refereed

Keywords

  • CNN
  • Cloud types
  • Deep learning
  • Image classification
  • Transfer learning

Fingerprint

Dive into the research topics of 'Cloud type classification using deep learning with cloud images'. Together they form a unique fingerprint.

Cite this