RAME Logo

RAME PUBLISHERS

A better space for Quality Research
International Journal of Computational and Electronics Aspects in Engineering (IJCEAE)
Volume 7 · Issue 5 · November 2026 · pp. 604–615
Research Article · Peer Reviewed
Received: August 25, 2026 · Accepted: October 20, 2026 · Published: November 7, 2026
DOI: 10.26706/ijceae.7.5.20261006
Open Access · CC BY 4.0

Age Category Classification Based on Hybrid CNN-Based Architecture with a Transformer-Inspired Feature-Transformation Module

Neam Salim M. Sheet*
Polytechnic College of Mosul, Northern Technical University, Mosul, Iraq.
*Correspondence: mti.lec40.neam@ntu.edu.iq
Download PDF Cite this article Certificate Volume 7 Issue 5

Abstract

The age-group classification of face images is very important in the computer vision field and has many applications in healthcare, security, and computer-human interaction. This paper aims to propose an architecture for mixed deep learning that combines the strengths of convolutional neural networks (CNN) and transformer-inspired multi-layer perceptron blocks for exact age-group classification. Our proposed model uses convolutional layers to extract the hierarchical local facial features and transformer-inspired multi-layer perceptron blocks that follow it which is responsible for refining the resulting feature vector through nonlinear fully connected transformation before final classification.

The model classifies the images of faces into four groups: infancy, childhood, youth, and elderly. Each subset of the UTKFace dataset was constructed in a balanced way and contained 1,800 images per class (7,200 images in total). To avoid the bias toward any class, the network was implemented and trained using MATLAB R2022b. The suggested architecture attains an error rate of 10.83% and an overall test accuracy of 89.17%. The highest precision was obtained for the infancy class (0.955), while the lowest was for the youth class (0.851). This hybrid model shows a more robust performance and better generalization than CNN models. The combination of convolutional layers with a transformer-inspired multi-layer perceptron block is an active approach for age-related visual recognition and offers a reliance solution for real-world applications. Note that confusion in class detection is possible because the real guessing by humans may also lead to age error, so what about the neural network model that is trained by humans.

Keywords

Age-group classification multi-layer perceptron convolutional neural networks Deep learning UTKFace

References

  1. M. Sowmya and S. R. Dutta, “NeuroAgeFusionNet an ensemble deep learning framework integrating CNN, transformers, and GNN for robust brain age estimation using MRI scans,” Sci. Rep., vol. 15, Art. no. 41996, 2025.
  2. N. Z. Waisi, N. Y. Abdullah, and M. T. Ghazal, “The automatic detection of underage troopers from live-videos based on deep learning,” Przegląd Elektrotechniczny, vol. 97, no. 9, pp. 85–88, 2021, doi: 10.15199/48.2021.09.18.
  3. L. Meng et al., “AdaViT: Adaptive vision transformers for efficient image recognition,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), New Orleans, LA, USA, Jun. 2022, pp. 12309–12318.
  4. K. Djoumessi, S. Ofosu Mensah, and P. Berens, “A hybrid fully convolutional CNN-transformer model for inherently interpretable medical image classification,” arXiv preprint arXiv:2504.08481, 2025.
  5. W. Tanveer, L. Fernández-Robles, E. Fidalgo, V. González-Castro, E. Alegre, and M. Mirjalili, “Comparative analysis of CNNs and vision transformers for age estimation,” in Proc. XLVI Jornadas de Automática, Cartagena, Spain, Sep. 2025, doi: 10.17979/ja-cea.2025.46.12251.
  6. M. Alrubaye, A. A. Hameed, and A. Jamil, “Enhancing human age detection: The impact of data augmentation and balancing on CNN performance,” in Proc. Int. Conf. Intell. Syst., Blockchain, Commun. Technol. (ISBCom), Lecture Notes in Networks and Systems, vol. 1268, Cham, Switzerland: Springer, 2025, pp. 803–818, doi: 10.1007/978-3-031-82377-0_64.
  7. G. George, S. A. Adeshina, and M. J. Haruna, “Chronological age estimation using multi-task convolutional neural networks with facial images and weight sharing,” in Proc. 3rd Int. Conf. Frontiers Academic Res. (ICFAR), Konya, Turkey, Jun. 2024.
  8. G. George, S. Adeshina, and M. M. Boukar, “Age estimation from facial images using custom convolutional neural network (CNN),” in Proc. 1st Int. Conf. Frontiers Academic Res. (ICFAR), Konya, Turkey, 2023, vol. 1.
  9. Y. Wang, Y. Deng, Y. Zheng, P. Chattopadhyay, and L. Wang, “Vision transformers for image classification: A comparative survey,” Technologies, vol. 13, no. 1, Art. no. 32, Jan. 2025, doi: 10.3390/technologies13010032.
  10. M. Khalil, A. Khalil, and A. Ngom, “A comprehensive study of vision transformers in image classification tasks,” arXiv preprint arXiv:2312.01232, 2023.
  11. Y. Huo, K. Jin, J. Cai, H. Xiong, and J. Pang, “Vision transformer (ViT)-based applications in image classification,” in Proc. IEEE 9th Int. Conf. Big Data Secur. Cloud (BigDataSecurity), New York, NY, USA, May 2023, pp. 135–140.
  12. A. Singh and V. K. Singh, “A hybrid transformer-sequencer approach for age and gender classification from in-the-wild facial images,” Neural Comput. Appl., vol. 36, no. 3, pp. 1149–1165, 2024, doi: 10.1007/s00521-023-09087-7.
  13. S. Sarıateş and E. Özbay, “Transformer-based multi-task learning with a task-aware weighted ensemble for simultaneous age estimation and gender classification from facial images,” Sensors, vol. 26, no. 18, Art. no. 5982, Sep. 2026, doi: 10.3390/s26185982.
  14. A. R. H. Al-Jader, N. S. M. Sheet, M. N. Farhan, and R. R. O. Al-Nima, “Normalized deep learning model for classifying white blood cell,” J. Image Graph., vol. 14, no. 2, pp. 259–266, Mar. 2026, doi: 10.18178/joig.14.2.259-266.
  15. M. Duan, K. Li, C. Yang, and K. Li, “A hybrid deep learning CNN–ELM for age and gender classification,” Neurocomputing, vol. 275, pp. 448–461, 2018.
  16. X. Zhang, “A hybrid CNN-transformer architecture for image classification on small datasets,” in Proc. 5th Int. Conf. Sensors Inf. Technol. (ICSI), Nanjing, China, Mar. 2025, pp. 324–331.
  17. A. H. Maray, M. N. Abed, N. S. M. Sheet, and R. A. Hussein, “Designing and implementing a fake document detection system using artificial intelligence,” J. Eng., vol. 2025, Art. no. 3902628, 2025.
  18. O. I. Alsaif, S. Q. Hasan, and A. H. Maray, “Using skeleton model to recognize human gait gender,” IAES Int. J. Artif. Intell., vol. 12, no. 2, pp. 974–983, Jun. 2023.
  19. K. Zhang et al., “Age group and gender estimation in the wild with deep RoR architecture,” IEEE Access, vol. 5, pp. 22492–22503, 2017.
  20. I. Naznin and M. M. Islam, “Age estimation from face image leveraging concatenated features of vision transformer along with ResNet-50,” in Proc. 26th Int. Conf. Comput. Inf. Technol. (ICCIT), Cox's Bazar, Bangladesh, Dec. 2023, pp. 1–6.
  21. S. Q. Alhashmi, K. H. Thanoon, and O. I. Alsaif, “A proposed face recognition based on hybrid algorithm for features extraction,” in Proc. 6th Int. Eng. Conf. Sustain. Technol. Develop. (IEC), Erbil, Iraq, Feb. 2020, pp. 232–236.
  22. R. Abdulhammed, “A secure AI-based thalassemia diagnosis framework,” Int. J. Comput. Electron. Aspects Eng. (IJCEAE), vol. 7, no. 4, p. 389, 2026.
  23. E. Agustsson, R. Timofte, S. Escalera, X. Baro, I. Guyon, and R. Rothe, “Apparent and real age estimation in still images with deep residual regressors on APPA-REAL database,” in Proc. 12th IEEE Int. Conf. Autom. Face Gesture Recognit. (FG), Washington, DC, USA, May/Jun. 2017, pp. 87–94.
  24. X. Chen, “Hybrid CNN-transformer architectures for medical image segmentation,” IEEE Access, vol. 11, pp. 117–128, 2023.
  25. H. H. Mohammed, S. A. Baker, and O. I. Alsaif, “An improved underwater image enhancement approach for border security,” J. Image Graph., vol. 12, pp. 199–204, 2024.
  26. M. N. Hussein, “Evolutionary CNN channel selection using online genetic algorithm: Brain tumor MRI as a case study,” Int. J. Comput. Electron. Aspects Eng. (IJCEAE), vol. 7, no. 4, p. 350, 2026.
  27. T. Y. Abdulqader, S. A. Abdulqader, and S. A. Baker, “Enhancing image recognition with quaternion neural networks: A novel approach to color layer integration,” J. Image Graph., vol. 13, no. 5, pp. 469–475, 2025, doi: 10.18178/joig.13.5.469-475.
  28. N. Y. Abdullah, M. T. Ghazal, and N. Waisi, “Pedestrian age estimation based on deep learning,” Indonesian J. Elect. Eng. Comput. Sci., vol. 22, pp. 1548–1555, 2021.
  29. D. Al-Obaidy and S. A. M. Dali, “Real-time face emotion recognition using deep learning with two-tier architectures for age and gender,” J. Methodical Comput. Commun., vol. 2025, pp. 33–40, 2025.
  30. M. K. Yousif, “Enhanced diabetic retinopathy classification using hybrid EfficientNetB0 with CBAM,” Int. J. Comput. Electron. Aspects Eng. (IJCEAE), vol. 7, no. 2, p. 176, 2026.
  31. H. A. Saleh and O. I. Alsaif, “A review on EEG-based emotion recognition: Methods, challenges, and future directions,” Int. J. Comput. Electron. Aspects Eng. (IJCEAE), vol. 7, no. 2, p. 278, 2026.
  32. S. K. Ghanim, “Hybrid ResNet50 and DenseNet121 with adaptive weighted feature fusion in infectious chest diseases classification,” Int. J. Comput. Electron. Aspects Eng. (IJCEAE), vol. 7, no. 2, p. 211, 2026.
  33. S. M. Amen, I. R. Alkhayyat, and H. O. Mahmood, “XAI-SecEx: An explainable deep learning framework for network intrusion detection,” Int. J. Comput. Electron. Aspects Eng. (IJCEAE), vol. 7, no. 2, p. 200, 2026.
  34. R. H. Ali Al-Mallah, “A novel hybrid InceptionV3-SqueezeNet network (HISNet) model for kidney disease classification in CT medical image,” Int. J. Comput. Electron. Aspects Eng. (IJCEAE), vol. 7, no. 4, p. 338, 2026.