The Influence of Activation Functions in Deep Learning Models Using Transfer Learning for Facial Age Prediction


  • Gilbert George Department of Computer Science, Baze University, Abuja, Nigeria
  • Steve Adeshina Department of Engineering, Nile University, Abuja, Nigeria.
  • Moussa Mahamat Boukar Department of Computer Science, Nile University, Abuja, Nigeria.


Activation Function, facial age detection, Transfer Learning, Regression, Deep learning, Convolutional Neural Networks (CNNs).


Due to its superior performance, the convolutional neural network (CNN) has been extensively applied in image recognition. A facial age prediction technique based on the CNN model is suggested in this research, due to its many uses in the sports industry, access control and age verification systems. The activation function is at the center of the CNN model's complicated hierarchical structure since it has the nonlinear properties that give the deep neural network its accurate artificial intelligence. The ReLu function is one of the best common activation functions, however, it has flaws. It is likely to manifest as the phenomena of neuronal necrosis since the derivative of the ReLu function is always zero when the input value is negative. The impact of the activation function in the CNN model is also examined in this research to address the aforementioned issue. We look at both linear and nonlinear activation functions for facial age prediction using three facial datasets, namely UTK Facial, IMDB-WIKI and CASIA African Facial datasets. In facial age prediction tasks built on the Keras framework, we have investigated Linear and ReLu activation functions in the output layer on four CNN architectures namely VGG16, VGG19, ResNet50 and MobileNet. we compared each model using the two activation functions in the last output dense layer. The experimental results show that MobileNet using ReLu Performed the best with a Mean Absolute Error of 1.03 on the CASIA African Facial Dataset and VGG16 with a Mean Absolute Error of 1.75 using the linear activation function on the UTK Facial dataset, we also demonstrate that the convolutional neural network based on the modified activation function outperforms most state-of-the-art activation models in terms of performance.


Download data is not yet available.


“VGGNet-16 Architecture. A Complete Guide | Kaggle.” https.// (accessed May 07, 2023).

S. Mascarenhas, M. A.-2021 I. C. on, and undefined 2021, “A comparison between VGG16, VGG19 and ResNet50 architecture frameworks for Image Classification,”, Accessed. May 07, 2023. [Online]. Available. https.//

Y. Wang, Y. Li, Y. Song, and X. Rong, “The influence of the activation function in a convolution neural network model of facial expression recognition,” Applied Sciences (Switzerland), vol. 10, no. 5, 2020, doi. 10.3390/app10051897.

J. Prajapati, A. Patel, and P. Raninga, “Facial Age Group Classification,” IOSR Journal of Electronics and Communication Engineering, vol. 9, no. 1, pp. 33–39, 2014, doi. 10.9790/2834-09123339.


M. F. Mustapha, N. M. Mohamad, G. Osman, and S. H. A. Hamid, “Age group classification using Convolutional Neural Network (CNN),” in Journal of Physics. Conference Series, 2021. doi. 10.1088/1742-6596/2084/1/012028.

V. Raman, K. Elkarazle, and P. Then, “Gender-specific Facial Age Group Classification Using Deep Learning,” Intelligent Automation and Soft Computing, vol. 34, no. 1, 2022, doi. 10.32604/iasc.2022.025608.

A. Tunc, S. Tasdemir, M. Koklu, and A. C. Cinar, “Age group and gender classification using convolutional neural networks with a fuzzy logic-based filter method for noise reduction,” Journal of Intelligent and Fuzzy Systems, vol. 42, no. 1, 2022, doi. 10.3233/JIFS-2191206.

J. Muhammad, Y. Wang, C. Wang, K. Zhang, Z. Sun, and S. Member, “CASIA-Face-Africa. A Large-scale African Face Image Database”, Accessed. Jul. 04, 2023. [Online]. Available. http.//

M. S. Islam, I. Hussain, M. M. Rahman, S. J. Park, and M. A. Hossain, “Explainable Artificial Intelligence Model for Stroke Prediction Using EEG Signal,” Sensors (Basel), vol. 22, no. 24, Dec. 2022, doi. 10.3390/s22249859.

D. Gunning, M. Stefik, J. Choi, T. Miller, S. Stumpf, and G. Z. Yang, “XAI—Explainable artificial intelligence,” Sci Robot, vol. 4, no. 37, Dec. 2019, doi. 10.1126/SCIROBOTICS.AAY7120.

K. Mohammed and G. George, “IDENTIFICATION AND MITIGATION OF BIAS USING EXPLAINABLE ARTIFICIAL INTELLIGENCE (XAI) FOR BRAIN STROKE PREDICTION,” Open Journal of Physical Science (ISSN. 2734-2123), vol. 4, no. 1, pp. 19–33, Apr. 2023, doi. 10.52417/OJPS.V4I1.457.

V. Sheoran, S. Joshi, and T. R. Bhayani, “Age and Gender Prediction Using Deep CNNs and Transfer Learning,” in Communications in Computer and Information Science, 2021. doi. 10.1007/978-981-16-1092-9_25.

B. S. Anami and C. V. Sagarnal, “Influence of Different Activation Functions on Deep Learning Models in Indoor Scene Images Classification,” Pattern Recognition and Image Analysis, vol. 32, no. 1, 2022, doi. 10.1134/S1054661821040039.

L. Su, “Intelligent Strategy by Deep Learning for Thyroid Case Images Classification,” in Proceedings - 2020 13th International Conference on Intelligent Computation Technology and Automation, ICICTA 2020, 2020. doi. 10.1109/ICICTA51737.2020.00076.

J. Li, C. Nanchang, and K. Song, “Research on image classification based on deep learning,” in Proceedings - 20th IEEE/ACIS International Summer Conference on Computer and Information Science, ICIS 2021-Summer, 2021. doi. 10.1109/ICIS51600.2021.9516872.

S. Kothari, S. Deshmukh, and S. Mehta, “Comparison of Age, Gender and Ethnicity Prediction Using Traditional CNN and Transfer Learning,” in 2022 13th International Conference on Computing Communication and Networking Technologies, ICCCNT 2022, 2022. doi. 10.1109/ICCCNT54827.2022.9984552.

M. A. H. Akhand, Md. Ijaj Sayim, S. Roy, and N. Siddique, “Human Age Prediction from Facial Image Using Transfer Learning in Deep Convolutional Neural Networks,” pp. 217–229, 2020, doi. 10.1007/978-981-15-3607-6_17.




How to Cite

George, G. ., Adeshina, S. ., & Boukar , M. M. . (2023). The Influence of Activation Functions in Deep Learning Models Using Transfer Learning for Facial Age Prediction. International Journal of Intelligent Systems and Applications in Engineering, 12(2), 328–335. Retrieved from



Research Article