Klasifikasi Gambar Asli dan Buatan AI Menggunakan SE-ResNet18 dan SE-ResNet50

Aurelia Catherine Lay; Yuni Yamasari

doi:10.33506/insect.v12i2.5989

Authors

Aurelia Catherine Lay Program Studi Teknik Informatika, Universitas Negeri Surabaya, Surabaya
Yuni Yamasari Program Studi Teknik Informatika, Universitas Negeri Surabaya, Surabaya

DOI:

https://doi.org/10.33506/insect.v12i2.5989

Keywords:

Deteksi Gambar, SE-ResNet, Class Weight, Optimasi Hyperparameter Bayesian

Abstract

Perkembangan model generatif, seperti Generative Adversarial Networks (GAN) dan Diffusion Models, yang menghasilkan gambar berkualitas tinggi buatan AI yang sulit dibedakan dari gambar asli, menjadi pendorong utama penelitian ini. Tujuan penelitian ini adalah menggunakan arsitektur SE-ResNet50 dan SE-ResNet18 untuk membuat model klasifikasi yang mampu membedakan antara gambar asli dan gambar buatan AI. Teknik class weight dan optimasi hyperparameter yang memanfaatkan Optimasi Bayesian digunakan untuk mengatasi ketidakseimbangan data. Dataset Chameleon, yang berisi 26.033 gambar—11.170 gambar buatan AI dan 14.863 gambar asli digunakan. Berdasarkan hasilnya, SE-ResNet50 menunjukkan kemampuan generalisasi yang unggul dan mencapai akurasi 88,33%, presisi 91,60%, recall 87,58%, f1-score 0,88 ketika dipadukan dengan class weight dan Optimasi Bayesian. SE-ResNet18 mencapai akurasi 88,75%, presisi 0,9095%, recall 0,8917%, F1-score 0,88% pada skenario tanpa class weight dan Optimasi Bayesian. Penelitian ini menunjukkan bahwa arsitektur berbasis Squeeze-and-Excitation, yang dipadukan dengan teknik penyeimbangan data dan Optimasi Hyperparameter, terbukti menjadi pendekatan yang handal untuk klasifikasi gambar buatan AI ketika data tidak seimbang.

References

[1] S. Yan et al., “A Sanity Check for AI-generated Image Detection,” Feb. 2025, [Online]. Available: http://arxiv.org/abs/2406.19435

[2] Y. Wang, Y. Hao, and A. X. Cong, “Harnessing Machine Learning for Discerning AI-Generated Synthetic Images,” May 2024, [Online]. Available: http://arxiv.org/abs/2401.07358

[3] V. Dan, “Deepfakes as a Democratic Threat: Experimental Evidence Shows Noxious Effects That Are Reducible Through Journalistic Fact Checks,” International Journal of Press/Politics, vol. 31, no. 3, pp. 525–550, Sep. 2019, doi: 10.1177/19401612251317766.

[4] J. Tan, A. E. Cinar, and K. Benyekhlef, “Can You Trust What You See? Human and AI Detection of Synthetic Legal Evidence,” 2026.

[5] “Internet Crime Complaint Center (IC3) | Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud.” Accessed: Aug. 17, 2026. [Online]. Available: https://www.ic3.gov/PSA/2024/PSA241203

[6] W. Chen, K. Yang, Z. Yu, Y. Shi, and C. L. P. Chen, “A survey on imbalanced learning: latest research, applications and future directions,” Artif. Intell. Rev., vol. 57, no. 6, Jun. 2024, doi: 10.1007/s10462-024-10759-6.

[7] W. Hussain et al., “Ensemble genetic and CNN model-based image classification by enhancing hyperparameter tuning,” Sci. Rep., vol. 15, no. 1, Dec. 2025, doi: 10.1038/s41598-024-76178-3.

[8] A. R. Gunukula, H. Das Gupta, and V. S. Sheng, “Detecting AI-Generated Images Using a Hybrid ResNet-SE Attention Model,” Applied Sciences (Switzerland), vol. 15, no. 13, Jul. 2025, doi: 10.3390/app15137421.

[9] J. J. Bird and A. Lotfi, “CIFAKE: Image Classification and Explainable Identification of AI-Generated Synthetic Images,” Mar. 2023, [Online]. Available: http://arxiv.org/abs/2303.14126

[10] K. A. Abdal Mashkoor, “Classifying Real and AI-Generated Images Using Fine-Tuned ResNet50,” Journal of Advanced Artificial Intelligence, Engineering and Technology, vol. 01, no. 02, 2025, doi: 10.56147/aaiet.1.2.7.

[11] A. Asare, A. Adjei Broni, A. K. Asare Dickson, M. Sagoe, and J. M. Cudjoe, “Performance of ResNet-18 and InceptionResNetV2 in Automated Detection of Diabetic Retinopathy,” Medicine Advances, vol. 3, no. 3, pp. 231–241, Sep. 2025, doi: 10.1002/med4.70023.

[12] M. Subramanian et al., “Diagnosis of Retinal Diseases Based on Bayesian Optimization Deep Learning Network Using Optical Coherence Tomography Images,” Comput. Intell. Neurosci., vol. 2022, 2022, doi: 10.1155/2022/8014979.

[13] L. Alzubaidi et al., “Review of deep learning: concepts, CNN architectures, challenges, applications, future directions,” J. Big Data, vol. 8, no. 1, Dec. 2021, doi: 10.1186/s40537-021-00444-8.

[14] N. D. Pham and N. Chilamkurti, “Data Leakage Threats and Protection in Split Learning: A Survey,” Association for Computing Machinery (ACM), Dec. 2023, pp. 141–147. doi: 10.1145/3659154.3659189.

[15] H. Mo, Z. Kuang, H. Wang, X. Cai, and K. Cheng, “Enhancing Burn Diagnosis through SE-ResNet18 and Confidence Filtering,” Journal of Imaging Informatics in Medicine, Feb. 2025, doi: 10.1007/s10278-025-01495-6.

[16] M. Khan and Y. Hossni, “A comparative analysis of LSTM models aided with attention and squeeze and excitation blocks for activity recognition,” Sci. Rep., vol. 15, no. 1, Dec. 2025, doi: 10.1038/s41598-025-88378-6.

[17] W. Setiawan, M. Koeshardianto, E. M. S. Rochman, A. Rachmad, and T. Herawan, “Class Weighting in Imbalanced Data for Leukocyte Classification using Fine-Tuning Inception-v3,” Iraqi Journal for Computer Science and Mathematics, vol. 6, no. 3, pp. 380–395, 2025, doi: 10.52866/2788-7421.1283.

[18] M. Abdelhamid and A. Desai, “Balancing the Scales: A Comprehensive Study on Tackling Class Imbalance in Binary Classification,” Sep. 2024, [Online]. Available: http://arxiv.org/abs/2409.19751

[19] M. Masum et al., “Bayesian Hyperparameter Optimization for Deep Neural Network-Based Network Intrusion Detection.”

[20] D. Chicco and G. Jurman, “The Matthews correlation coefficient (MCC) should replace the ROC AUC as the standard metric for assessing binary classification,” BioData Min., vol. 16, no. 1, Dec. 2023, doi: 10.1186/s13040-023-00322-4.

[21] D. Stojnev and A. Stojnev Ilić, “Preprocessing Image Data for Deep Learning,” Singidunum University, Oct. 2020, pp. 312–317. doi: 10.15308/sinteza-2020-312-317.

[22] M. H. Guo et al., “Attention mechanisms in computer vision: A survey,” Sep. 01, 2022, Tsinghua University. doi: 10.1007/s41095-022-0271-y.

[23] I. Araf, A. Idri, and I. Chairi, “Cost-sensitive learning for imbalanced medical data: a review,” Artif. Intell. Rev., vol. 57, no. 4, Apr. 2024, doi: 10.1007/s10462-023-10652-8.

[24] J. Chai, H. Zeng, A. Li, and E. W. T. Ngai, “Deep learning in computer vision: A critical review of emerging techniques and application scenarios,” Machine Learning with Applications, vol. 6, p. 100134, 2021, doi: 10.24433/CO.0411648.v1.