The Effect of Activation Functions on Performance in Image Classification with Transfer Deep Learning Techniques

Year 2024, Volume: 24 Issue: 2, 294 - 307, 29.04.2024

Gül Cihan Habek Sakir Tasdemır Fatih Basciftci Ahmet Yılmaz

https://doi.org/10.35414/akufemubid.1334098

Abstract

Convolutional neural networks (CNN), a feedforward neural network model, are frequently used in image classification problems. New architectures have been developed by making some additions and changes in order to improve the performance on the CNN model, which generally consists of convolution layer, pooling layer and fully connected layer. By adding different numbers of hidden layers with sigmoid, tanh and ReLu activation functions to the CNN-based VGG16 and ResNet50 architectures developed in this study, images were classified by deep transfer learning technique and the performances were compared. The dataset used for classification is a multi-class dataset related to fruits, and a 10-neuron softmax classifier was used in the last layer of the established models. Entering the epoch as 10, results were obtained for four different metrics: accuracy, precision, recall, and f1-score. When the results are compared, it has been observed that the model created by adding a total of three layers, including two hidden layers with 256 and 128 neurons, and a softmax classifier layer with 10 neurons, to the ResNet50 architecture with sigmoid activation function, gives the best result with a classification accuracy value of %97.5. Finally, the results obtained for the four metrics were subjected to the Friedman and Nemenyi post-hoc tests, a statistical analysis was made and the relationship between the models was tested. It was concluded that the models created as a result of the test were related to each other.

Keywords

Deep Learning, ResNet50 Architecture, Transfer Deep Learning, VGG16 Architecture, Image Classification

Transfer Derin Öğrenme Teknikleri ile Görüntü Sınıflandırmada Aktivasyon Fonksiyonlarının Performans Üzerindeki Etkisi

Abstract

İleri beslemeli yapay sinir ağı modeli olan konvolüsyonel sinir ağları (CNN) görüntülerin sınıflandırılması problemlerinde sıklıkla kullanılmaktadır. Genel olarak konvolüsyon katmanı, havuzlama katmanı ve tam bağlı katmandan oluşan CNN modeli üzerinde performansı iyileştirmek amacı ile birtakım eklemeler ve değişiklikler yapılarak yeni mimariler geliştirilmiştir. Bu çalışmada geliştirilen CNN tabanlı VGG16 ve ResNet50 mimarilerine sigmoid, tanh ve ReLu aktivasyon fonksiyonlu farklı sayıda gizli katman eklenerek derin transfer öğrenme tekniği ile görüntüler sınıflandırılmış ve performansları karşılaştırılmıştır. Sınıflandırma için kullanılan veri seti meyveler ile ilgili çoklu sınıflı bir veri seti olup kurulan modellerin son katmanında 10 nöronlu softmax sınıflandırıcı kullanılmıştır. Devir sayısı 10 girilerek sınıflandırma doğruluğu (accuracy), duyarlılık (precision), geri çağırma (recall) ve f1-ölçütü olmak üzere dört farklı metrik için sonuçlar alınmıştır. Alınan sonuçlar kıyaslandığında modeller arasında ResNet50 mimarisine sigmoid aktivasyon fonksiyonlu, 256 ve 128 nöronlu iki gizli katman ve 10 nöronlu bir softmax sınıflandırıcı katmanı olmak üzere toplam üç katman eklenerek oluşturulan modelin %97.5 sınıflandırma doğruluğu değeri ile en iyi sonucu verdiği gözlemlenmiştir. Son olarak dört metrik için alınan sonuçlar Friedman ve Nemenyi post-hoc testlerine tabi tutularak istatistiksel bir analiz yapılmış, modeller arasındaki ilişki test edilmiştir. Test sonucunda oluşturulan modellerin birbirleri ile ilişkili olduğu sonucuna varılmıştır.

Keywords

Görüntü Sınıflandırma, Derin Öğrenme, ResNet50 Mimarisi, Transfer Derin Öğrenme, VGG16 Mimarisi

References

• Alkhouly, A., Mohammed, A. and Hefny, Hesham, H., 2021, Improving The Performance Of Deep Neural Networks Using Two Proposed Activation Functions, IEEE Access, 1-1. https://doi.org/10.1109/ACCESS.2021.3085855
• Al-Saedi, D.K.A. and Savaş, S., 2022. Classification of Skin Cancer with Deep Transfer Learning Method, Computer Science. IDAP-2022, International Artificial Intelligence and Data Processing Symposium, 202-210. https://doi.org/10.53070/bbd.1172782
• Ammatmanee, C. and Gan, L., 2021. Transfer learning for hostel image classification. Data Technologies and Applications, 56, 44-59. https://doi.org/10.1108/DTA-02-2021-0042
• Behera, S.K., Rath, A.K. and Sethy, P.K., 2021. Maturity status classification of papaya fruits based on machine learning and transfer learning approach. Information Processing in Agriculture, 8, 2, 244-250. https://doi.org/10.1016/j.inpa.2020.05.003
• Bozkurt, F., 2021/1, Derin Öğrenme Tekniklerini Kullanarak Akciğer X-Ray Görüntülerinden COVID-19 Tespiti, Avrupa Bilim Ve Teknoloji Dergisi, (24), 149-156. https://doi.org/10.31590/ejosat.898385
• Bozkurt, F., 2021/2, A Study on CNN Based Transfer Learning for Recognition of Flower Species, 2021, European Journal of Science and Technology, 32, 883-890. https://doi.org/10.31590/ejosat.1039632
• Bozkurt, F., 2022, A Deep And Handcrafted Features-Based Framework For Diagnosis Of COVID-19 From Chest X-Ray Images, Concurrency Computat Pract Exper, 34(5). https://doi.org/10.1155/2021/6799202
• Buchanan, B.G., 2005. A (Very) Brief History of Artificial Intelligence. AI Magazine, 26, 4, 53-60. https://doi.org/10.1609/aimag.v26i4.1848
• Chen, Y., Lin, Y., Xu, X., Ding, J., Li, C., Zeng, Y., Liu, W., Xie, W. and Huang, J., 2022. Classification of lungs infected COVID-19 images based on inception-ResNet. Computer Methods and Programs in Biomedicine, 225, 1-9. https://doi.org/10.1016/j.cmpb.2022.107053
• Doğan, F., ve Türkoğlu, İ., 2019. Derin Öğrenme Modelleri ve Uygulama Alanlarına İlişkin Bir Derleme. Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi, 10, 2, 409-445. https://doi.org/10.24012/dumf.411130
• Dubey, S. R., Singh, S. K. and Chaudhuri, B. B., 2022, Activation Functions in Deep Learning: A Comprehensive Survey and Benchmark, arXiv. https://doi.org/10.48550/arXiv.2109.14545
• Friedman, M.A., 1940. Comparison of alternative tests of significance for the problem of m rankings. Annals of Mathematical Statistics, 11, 1, 86-92. https://doi.org/10.1214/aoms/1177731944
• Glorot, X., Bordes, A. and Bengio, Y., 2011, Deep sparse rectifier neural networks, In Proceedings of the fourteenth international conference on artificial intelligence and statistics, pages, 315–323.
• Gulzar, Y., 2023. Fruit Image Classification Model Based on MobileNetV2 with Deep Transfer Learning Technique. Sustainability, 15, 3, 1906. https://doi.org/10.3390/su15031906
• Habek, G.C., 2022. Makine Öğrenmesi Teknikleriyle Kripto Para Duygu Analizi, Yüksek Lisans Tezi, Manisa Celal Bayar Üniversitesi Fen Bilimleri Enstitüsü, Manisa, 75.
• Hao, W., Yizhou, W., Yaqin, L. and Zhili, S., 2020, The Role of Activation Function in CNN, 2020 2nd International Conference on Information Technology and Computer Application (ITCA), Guangzhou, China, 429-432. https://doi.org/10.1109/ITCA52113.2020.00096
• He, K., Zhang, X., Ren, S. and Sun, J., 2015. Deep Residual Learning for Image Recognition. arXiv. https://doi.org/10.48550/arXiv.1512.03385
• Hemalatha, N., Sukhetha, P. ve Sukumar, R., 2022. Classification of Fruits and Vegetables using Machine and Deep Learning Approach. In 2022 International Conference on Trends in Quantum Computing and Emerging Business Technologies (TQCEBT), 1-4. https://doi.org/10.1109/TQCEBT54229.2022.10041654
• Mascarenhas, S. and Agarwal, M., 2021. A comparison between VGG16, VGG19 and ResNet50 architecture frameworks for Image Classification. 2021 International Conference on Disruptive Technologies for Multi-Disciplinary Research and Applications (CENTCON-2021), Bengaluru-India, 96-99. https://doi.org/10.1109/CENTCON52345.2021.9687944
• Memiş, S., Enginoğlu, S. ve Erkan, U., 2022. A new classification method using soft decision-making based on an aggregation operator of fuzzy parameterized fuzzy soft matrices. Turkish Journal of Electrical Engineering and Computer Sciences, 30, 3, 871-890. https://doi.org/10.55730/1300-0632.3816
• Mukhiddinov, M., Muminov. A. and Cho, J., 2022. Improved Classification Approach for Fruits and Vegetables Freshness Based on Deep Learning, Sensors, 22, 21, 1-20. https://doi.org/10.3390/s22218192
• Nemenyi, P.B., 1963. Distribution-free multiple comparisons. PhD, Princeton University, Princeton, New Jersey, USA.
• Nwankpa, C., Ijomah, W., Gachagan, A. and Marshall, S., 2018, Activation Functions: Comparison of trends in Practice and Research for Deep Learning, arXiv. https://doi.org/10.48550/arXiv.1811.03378
• Özçelik, Y. B. ve Altan, A., 2021. Diyabetik Retinopati Teşhisi için Fundus Görüntülerinin Derin Öğrenme Tabanlı Sınıflandırılması. Avrupa Bilim Ve Teknoloji Dergisi(29), 156-167. https://doi.org/10.31590/ejosat.1011806
• Ponce, J.M., Aquino, A. and Andújar, M., 2019. Olive-Fruit Variety Classification by Means of Image Processing and Convolutional Neural Networks. IEEE Access, 147629-147641. https://doi.org/10.1109/ACCESS.2019.2947160
• Rojas-Aranda, J.L., Nunez-Varela, J.I., Cuevas-Tello, J.C. and Rangel-Ramirez, G., 2020. Fruit Classification for Retail Stores Using Deep Learning. 12th Mexican Conference on Pattern Recognition (MCPR), 3-13. https://doi.org/10.1007/978-3-030-49076-8_1
• Simonyan, K. and Zisserman, A., 2014. Very Deep Convolutional Networks for Large-Scale Image Recognition. arXiv. https://doi.org/10.48550/arXiv.1409.1556
• Talo, M., 2019. Automated classification of histopathology images using transfer learning. Artificial Intelligence in Medicine, 101, 1-8. Probabilistic losses, https://keras.io/api/losses/probabilistic_losses, (18.12.2023)