Binary Image Classification with CNNs, Transfer Learning and Classical Models
, Dimitrios Vasileios Oikonomou2, Sofia Panagiota Chaliasou3 and Nikolaos Rigas4(This article belongs to the Special Issue on Special Issue on Multidisciplinary Sciences and Advanced Technology (SI-MSAT 2025) and the Section Artificial Intelligence – Computer Science (AIC))
Export Citations
Cite
Oikonomou, N. V. , Oikonomou, D. V. , Chaliasou, S. P. and Rigas, N. (2026). Binary Image Classification with CNNs, Transfer Learning and Classical Models. Journal of Engineering Research and Sciences, 5(1), 66–75. https://doi.org/10.55708/js0501006
Nikolaos Vasileios Oikonomou, Dimitrios Vasileios Oikonomou, Sofia Panagiota Chaliasou and Nikolaos Rigas. "Binary Image Classification with CNNs, Transfer Learning and Classical Models." Journal of Engineering Research and Sciences 5, no. 1 (January 2026): 66–75. https://doi.org/10.55708/js0501006
N.V. Oikonomou, D.V. Oikonomou, S.P. Chaliasou and N. Rigas, "Binary Image Classification with CNNs, Transfer Learning and Classical Models," Journal of Engineering Research and Sciences, vol. 5, no. 1, pp. 66–75, Jan. 2026, doi: 10.55708/js0501006.
This study presents a comprehensive comparative analysis of binary face classification utilizing Deep Learning and traditional Machine Learning approaches. We evaluate three distinct modeling strategies: (1) End-to-end Convolutional Neural Networks (CNNs), including a baseline TensorFlow model and an optimized PyTorch architecture; (2) Hybrid CNN-MLP networks; and (3) Feature extraction via a pre-trained ResNet50 coupled with classical classifiers (Random Forest, Logistic Regression). The experimental dataset consists of 6,376 face images (5,102 training, 1,274 validation) derived from a Kaggle challenge. We implement rigorous data augmentation (rotation, shifts, flips) and regularization techniques (Dropout, Batch Normalization, Weight Decay) to mitigate overfitting. Results demonstrate that the optimized PyTorch CNN achieved the highest generalization performance with a validation accuracy of ~85.9% and an AUC of 0.94, utilizing AdamW optimizer and Cosine Annealing scheduling. Conversely, the classical models (Random Forest, Logistic Regression) utilizing ResNet50 features exhibited near-perfect training metrics (AUC ≈ 1.0) and competitive validation accuracy (>90%), highlighting the efficacy of transfer learning. We critically analyze the "underfitting" phenomenon observed in the baseline CNN (Training Accuracy < Validation Accuracy) attributing it to aggressive regularization. This work provides a clear roadmap for selecting between computational-heavy deep architectures and efficient feature-based classical models based on available resources and accuracy requirements.
- A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” in Advances in Neural Information Processing Systems, 2012, pp. 1097–1105. doi:10.1145/3065386.
- K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778. doi:10.1109/CVPR.2016.90.
- J. Yosinski, J. Clune, Y. Bengio, and H. Lipson, “How transferable are features in deep neural networks?,” in Advances in neural information processing systems, 2014, pp. 3320–3328. Link: https://proceedings.neurips.cc/paper/2014/file/375c71349b295f059961d99a30030c5e-Paper.pdf
- L. Breiman, “Random Forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001. doi:10.1023/A:1010933404324.
- S. Ioffe and C. Szegedy, “Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift,” arXiv preprint arXiv:1502.03167, 2015. doi:10.48550/arXiv.1502.03167.
- M. Tan and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” in International Conference on Machine Learning (ICML), 2019, pp. 6105–6114. Link: https://proceedings.mlr.press/v97/tan19a.html
- S. Yang, W. Xiao, M. Zhang, S. Guo, J. Zhao, and F. Shen, “Image Data Augmentation for Deep Learning: A Survey,” arXiv preprint arXiv:2204.08610, 2023. doi:10.48550/arXiv.2204.08610.
- D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” arXiv preprint arXiv:1412.6980, 2014. doi:10.48550/arXiv.1412.6980.
- Kaggle, “Face Classification Dataset,” Kaggle Datasets, [Online]. Available: https://www.kaggle.com/ (Accessed: 2024).
- R. Kohavi, “A study of cross-validation and bootstrap for accuracy estimation and model selection,” in Proceedings of the 14th International Joint Conference on Artificial Intelligence (IJCAI), 1995, vol. 2, pp. 1137–1143. Link: https://www.ijcai.org/Proceedings/95-2/Papers/016.pdf
- Y. LeCun, L. Bottou, G. B. Orr, and K.-R. Müller, “Efficient BackProp,” in Neural Networks: Tricks of the Trade, Springer, 2012, pp. 9–48. doi:10.1007/978-3-642-35289-8_3.
- C. Shorten and T. M. Khoshgoftaar, “A survey on Image Data Augmentation for Deep Learning,” Journal of Big Data, vol. 6, no. 1, p. 60, 2019. doi:10.1186/s40537-019-0197-0.
- L. Perez and J. Wang, “The Effectiveness of Data Augmentation in Image Classification using Deep Learning,” arXiv preprint arXiv:1712.04621, 2017. DOI: doi:10.48550/arXiv.1712.04621.
- M. Abadi et al., “TensorFlow: A System for Large-Scale Machine Learning,” in Proceedings of the 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2016, pp. 265–283. Link: https://www.usenix.org/system/files/conference/osdi16/osdi16-abadi.pdf
- A. Paszke et al., “PyTorch: An Imperative Style, High-Performance Deep Learning Library,” in Advances in Neural Information Processing Systems (NeurIPS), 2019, vol. 32, pp. 8024–8035. Link: https://proceedings.neurips.cc/paper/2019/file/bdbca288fee7f92f2bfa9f7012727740-Paper.pdf
- F. Pedregosa et al., “Scikit-learn: Machine Learning in Python,” Journal of Machine Learning Research, vol. 12, pp. 2825–2830, 2011. Link: https://jmlr.org/papers/v12/pedregosa11a.html
- N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: A Simple Way to Prevent Neural Networks from Overfitting,” The Journal of Machine Learning Research, vol. 15, no. 1, pp. 1929–1958, 2014. Link: https://jmlr.org/papers/v15/srivastava14a.html
- B. Xu, N. Wang, T. Chen, and M. Li, “Empirical Evaluation of Rectified Activations in Convolutional Network,” arXiv preprint arXiv:1505.00853, 2015. doi:10.48550/arXiv.1505.00853.
- I. Loshchilov and F. Hutter, “Decoupled Weight Decay Regularization,” in International Conference on Learning Representations (ICLR), 2019. Link: https://openreview.net/forum?id=Bkg6RiCqY7
- I. Loshchilov and F. Hutter, “SGDR: Stochastic Gradient Descent with Warm Restarts,” in International Conference on Learning Representations (ICLR), 2017. Link: https://openreview.net/forum?id=Skq89Scxx
- A. S. Razavian, H. Azizpour, J. Sullivan, and S. Maki, “CNN Features Off-the-Shelf: An Astounding Baseline for Recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2014, pp. 806–813. doi:10.1109/CVPRW.2014.122.
- D. W. Hosmer Jr., S. Lemeshow, and R. X. Sturdivant, Applied Logistic Regression, 3rd ed. Hoboken, NJ: Wiley, 2013. doi:10.1002/9781118548387.
- A. P. Bradley, “The use of the area under the ROC curve in the evaluation of machine learning algorithms,” Pattern Recognition, vol. 30, no. 7, pp. 1145–1159, 1997. doi:10.1016/S0031-3203(96)00142-2.
- D. Chicco and G. Jurman, “The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation,” BMC Genomics, vol. 21, no. 6, 2020. doi:10.1186/s12864-019-6413-7.