EfficientNet-B0 ile Yer Tabanlı Bulut Görüntülerinin Sınıflandırılması: CCSN Veri Kümesi Üzerine Bir Çalışma

Öz

Bulutlar, yeryüzünün %60’ından fazlasını kaplamakta olup, kısa dalga ve uzun dalga radyasyonunu değiştirerek hidrolojik döngü, iklim sistemi ve radyasyon dengesi üzerinde önemli bir rol oynamaktadır. Hava tahminlerinin doğruluğu; havacılık, deniz taşımacılığı, tarım, enerji ve çevresel izleme gibi pek çok sektör için kritik öneme sahiptir. Bu çalışmada, yere dayalı bulut görüntülerinin sınıflandırılması amacıyla EfficientNet-B0 mimarisi kullanılarak derin öğrenme tabanlı bir yaklaşım geliştirilmiştir. Orijinal haliyle 2.543 görüntü içeren Cirrus Cumulus Stratus Nimbus (CCSN) veri seti kullanıldığında modelin doğruluk oranı %53 seviyesinde kalmıştır. Ancak, her bir bulut sınıfı için görüntü sayısı veri artırma (augmentation) teknikleriyle 1000'e eşitlenerek veri seti dengelendiğinde modelin başarımında belirgin bir artış gözlenmiş ve doğruluk oranı %90.14’e ulaşmıştır. Elde edilen sonuçlar, EfficientNet-B0 mimarisinin veri dengesi sağlandığında bulut sınıflandırma görevinde etkili bir performans sergilediğini göstermekte; meteorolojik analiz, havacılık ve iklim gözlem uygulamaları için umut vadeden bir çözüm sunmaktadır.

Anahtar Kelimeler

• Bulut Sınıflandırma
• Derin Öğrenme
• Efficientnet-B0
• Veri Artırma
• Meteorolojik Analiz
• Yapay Zekâ

Proje Numarası

223M590

Classification of Ground-Based Cloud Images Using EfficientNet-B0: A Study on the CCSN Dataset

Öz

Clouds cover more than 60% of the Earth's surface and play an important role in the hydrological cycle, climate system, and radiation balance by altering shortwave and longwave radiation. The accuracy of weather forecasts is critical for many sectors, including aviation, maritime transport, agriculture, energy, and environmental monitoring. In this study, a deep learning-based approach was developed using the EfficientNet-B0 architecture for the classification of ground-based cloud images. When using the original Cirrus Cumulus Stratus Nimbus (CCSN) dataset, which contains 2543 images, the model's accuracy rate remained at 53%. However, when the number of images for each cloud class was balanced to 1,000 using data augmentation techniques, a significant increase in model performance was observed, with the accuracy rate reaching 90.14%. The results obtained demonstrate that the EfficientNet-B0 architecture delivers effective performance in cloud classification tasks when data balance is achieved, offering a promising solution for meteorological analysis, aviation, and climate observation applications.

Anahtar Kelimeler

• Cloud Classification
• Deep Learning
• Efficientnet-B0
• Data Augmentation
• Meteorological Analysis
• Artificial Intelligence

Destekleyen Kurum

TUBITAK

Proje Numarası

223M590

Teşekkür

This study was supported by The Scientific and Technological Research Council of Türkiye (TÜBİTAK) under the 1001 – Scientific and Technological Research Projects Funding Program (Project No: 223M590), and by Fırat University Scientific Research Projects Unit (FÜBAP) within the scope of the comprehensive research project numbered MF.25.122.

Kaynakça

1. Wei, D., Ge, F., Zhang, B., Zhao, Z., Li, D., Xi, L., ... & Wang, X. (2025). CloudViT: A Lightweight Ground-Based Cloud Image Classification Model with the Ability to Capture Global Features. Computers, Materials & Continua, 83(3).
2. Lee, H. J., Kang, J. E., & Kim, C. H. (2015). Forty-year (1971–2010) semiquantitative observations of visibility-cloud-precipitation in Korea and its implication for aerosol effects on regional climate. Journal of the Air & Waste Management Association, 65(7), 788–799. https://doi.org/10.1080/10962247.2015.1016633
3. Zhang, Jinglin, et al. "CloudNet: Ground‐based cloud classification with deep convolutional neural network." Geophysical Research Letters 45.16 (2018): 8665-8672.
4. Song, Qianqian, Zhihui Cui, and Pu Liu. "An Efficient Solution for Semantic Segmentation of Three Ground‐based Cloud Datasets." Earth and Space Science 7.4 (2020): e2019EA001040.
5. Roy, Roshan, et al. "Towards automatic transformer-based cloud classification and segmentation." NeurIPS 2021 workshop on tackling climate change with machine learning. Vol. 2021. 2021.
6. Xiafukaiti, Alifu, et al. "Application of Tensorized Neural Networks for Cloud Classification." arXiv preprint arXiv:2405.10946 (2024).
7. Li, Sheng, et al. "CloudDenseNet: Lightweight ground-based cloud classification method for large-scale datasets based on reconstructed DenseNet." Sensors 23.18 (2023): 7957.
8. Jiang, Y., Cheng, W., Gao, F., Zhang, S., Wang, S., Liu, C., & Liu, J. (2022). A cloud classification method based on a convolutional neural network for FY-4A satellites. Remote Sensing, 14(10), 2314.

9. Guzel, Mehmet, et al. "Cloud type classification using deep learning with cloud images." PeerJ Computer Science 10 (2024): e1779.
10. Li, X., Qiu, B., Cao, G., Wu, C., & Zhang, L. (2022). A novel method for ground-based cloud image classification using transformer. Remote Sensing, 14(16), 3978.
11. Nie, Y., Li, X., Paletta, Q., Aragon, M., Scott, A., & Brandt, A. (2022). Open-source ground-based sky image datasets for very short-term solar forecasting, cloud analysis and modeling: A comprehensive survey. arXiv preprint arXiv:2211.14709.
12. Zhang, L., Wei, W., Qiu, B., Luo, A., Zhang, M., & Li, X. (2022). A novel ground-based cloud image segmentation method based on a multibranch asymmetric convolution module and attention mechanism. Remote Sensing, 14(16), 3970.Li, Xiaotong, et al. "A novel method for ground-based cloud image classification using transformer." Remote Sensing 14.16 (2022): 3978.
13. Toğaçar, Mesut, and Burhan Ergen. "Classification of cloud images by using super resolution, semantic segmentation approaches and binary sailfish optimization method with deep learning model." Computers and Electronics in Agriculture 193 (2022): 106724.
14. Fang, Chunyao, et al. "Research on cloud recognition technology based on transfer learning." 2019 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC). IEEE, 2019.
15. Cynthia, E. P., Ismanto, E., Arifandy, M. I., Sarbaini, S., Nazaruddin, N., Manuhutu, M. A., & Akbar, M. A. (2022, December). Convolutional Neural Network and Deep Learning Approach for Image Detection and Identification. In Journal of Physics: Conference Series (Vol. 2394, No. 1, p. 012019). IOP Publishing.
16. Manzo, M., & Pellino, S. (2021). Voting in transfer learning system for ground-based cloud classification. Machine Learning and Knowledge Extraction, 3(3), 542-553.
17. Li, Z., Kong, H., & Wong, C. S. (2023). Neural network-based identification of cloud types from ground-based images of cloud layers. Applied Sciences, 13(7), 4470.
18. Suemitsu, K., Endo, S., & Sato, S. (2024). Classification of Rainfall Intensity and Cloud Type from Dash Cam Images Using Feature Removal by Masking. Climate, 12(5), 70.
19. Zhang, J. L., Liu, P., Zhang, F., & Song, Q. Q. ( 2018). CloudNet: Ground‐based cloud classification with deep convolutional neural network. Geophysical Research Letters, 45, 8665– 8672. https://doi.org/10.1029/2018GL077787