A DEEP LEARNING MODEL WITH A NEW LOSS FUNCTION  FOR AGE ESTIMATION

ARMAN I.  MOHAMMED; SARBAST H.  ALI; OMER MOHAMMED SALIH  HASSAN; SARDAR OMAR  SALIH

doi:10.26682/sjuod.2023.26.2.32

ARMAN I. MOHAMMED Dept. of Information Technology, Duhok Polytechnic University, Kurdistan Region-Iraq
SARBAST H. ALI Dept. of Information Technology, Duhok Polytechnic University, Kurdistan Region-Iraq
OMER MOHAMMED SALIH HASSAN Dept. of Information Technology, Duhok Polytechnic University, Kurdistan Region-Iraq
SARDAR OMAR SALIH Dept. of Web technology, Duhok Polytechnic University, Kurdistan Region-Iraq

DOI: https://doi.org/10.26682/sjuod.2023.26.2.32

Keywords: Age Estimation, Convolutional Neural Networks, Deep Learning, Loss Function, Mean Squared Error.

Abstract

Age estimation is a global challenge in the area of computer vision, as it depends on the facial features of the person. Recently, it has become an important approach for facial recognition problems and many other real-world applications. Accurate age estimation has the potential to improve decision-making processes in various industries and applications. It has been shown that the approach of Convolutional Neural Network (CNN) performs well for age estimation and promising results have been obtained by many researchers. However, the efficiency of CNN, to a great extent, is determined by the strength of the loss function. In this paper, a groundbreaking contribution is introduced, presenting a loss function that effectively calculates the disparity between the real and predicted age labels. Which is driven from Golden ratio and Mean Squared Error (MSE) functions. The proposed loss function is denoted by Golden Mean Squared Error (GMSE). A predesigned CNN is trained with UTKFace and FG-Net age datasets. According to the results, GMSE proved to operate better than preexisted loss functions. The MSE loss at epoch 25 was 51.34 and the GMSE loss at the same epoch was 3.15. At final round of training, the MSE loss was 6.56 and the GMSE loss was 1.58. The Mean Absolute Error (MAE) loss function was also used, but it couldn't lower the loss below 2 in the last epoch. Furthermore, the GMSE accuracy outperformed both MSE and MAE in the testing phase for both the UTKFace and FG-NET datasets. The GMSE loss function achieved better results than the MSE and MAE loss functions, indicating that it can save time and computations during the training process and provide better results at production phaseAge estimation is a global challenge in the area of computer vision, as it depends on the facial features of the person. Recently, it has become an important approach for facial recognition problems and many other real-world applications. Accurate age estimation has the potential to improve decision-making processes in various industries and applications. It has been shown that the approach of Convolutional Neural Network (CNN) performs well for age estimation and promising results have been obtained by many researchers. However, the efficiency of CNN, to a great extent, is determined by the strength of the lAge estimation is a global challenge in the area of computer vision, as it depends on the facial features of the person. Recently, it has become an important approach for facial recognition problems and many other real-world applications. Accurate age estimation has the potential to improve decision-making processes in various industries and applications. It has been shown that the approach of Convolutional Neural Network (CNN) performs well for age estimation and promising results have been obtained by many researchers. However, the efficiency of CNN, to a great extent, is determined by the strength of the loss function. In this paper, a groundbreaking contribution is introduced, presenting a loss function that effectively calculates the disparity between the real and predicted age labels. Which is driven from Golden ratio and Mean Squared Error (MSE) functions. The proposed loss function is denoted by Golden Mean Squared Error (GMSE). A predesigned CNN is trained with UTKFace and FG-Net age datasets. According to the results, GMSE proved to operate better than preexisted loss functions. The MSE loss at epoch 25 was 51.34 and the GMSE loss at the same epoch was 3.15. At final round of training, the MSE loss was 6.56 and the GMSE loss was 1.58. The Mean Absolute Error (MAE) loss function was also used, but it couldn't lower the loss below 2 in the last epoch. Furthermore, the GMSE accuracy outperformed both MSE and MAE in the testing phase for both the UTKFace and FG-NET datasets. The GMSE loss function achieved better results than the MSE and MAE loss functions, indicating that it can save time and computations during the training process and provide better results at production phaseoss function. In this paper, a groundbreaking contribution is introduced, presenting a loss function that effectively calculates the disparity between the real and predicted age labels. Which is driven from Golden ratio and Mean Squared Error (MSE) functions. The proposed loss function is denoted by Golden Mean Squared Error (GMSE). A predesigned CNN is trained with UTKFace and FG-Net age datasets. According to the results, GMSE proved to operate better than preexisted loss functions. The MSE loss at epoch 25 was 51.34 and the GMSE loss at the same epoch was 3.15. At final round of training, the MSE loss was 6.56 and the GMSE loss was 1.58. The Mean Absolute Error (MAE) loss function was also used, but it couldn't lower the loss below 2 in the last epoch. Furthermore, the GMSE accuracy outperformed both MSE and MAE in the testing phase for both the UTKFace and FG-NET datasets. The GMSE loss function achieved better results than the MSE and MAE loss functions, indicating that it can save time and computations during the training process and provide better results at production phase

Downloads

Download data is not yet available.

References

Agbo-Ajala, O., & Viriri, S. (2020). Deeply Learned Classifiers for Age and Gender Predictions of Unfiltered Faces. The Scientific World Journal, 2020, 1-12.
Al-azzawi, D. S. (2019). Human Age and Gender Prediction Using Deep Multi-Task Convolutional Neural Network. Journal of Southwest Jiaotong University, 54(4), 1–11.
Al-Shannaq, A., & Lamiaa, E. (2020). Age Estimation Using Specific Domain Transfer Learning. Jordanian Journal of Computers and Information Technology (JJCIT), 6(2), 122–139.
Berg, A., Oskarsson, M., & O’Connor, M. (2021). Deep Ordinal Regression with Label Diversity. 2020 25th International Conference on Pattern Recognition (ICPR), vol. 2, 2740–2747.
Bui, T., Ribeiro, L., Ponti, M., & Collomosse, J. (2018). Sketching out the details: Sketch-based image retrieval using convolutional neural networks with multi-stage regression. Computers & Graphics, 71, 77–87.
Dhimar, T., & Mistree, K. (2016, March). Feature extraction for facial age estimation: A survey. In 2016 International Conference on Wireless Communications, Signal Processing and Networking (WiSPNET) (pp. 2243-2248). IEEE.
Dornaika, F., Bekhouche, S. E., & Arganda-Carreras, I. (2020). Robust regression with deep CNNs for facial age estimation: An empirical study. Expert Systems with Applications, 141, 112942-112948.
Eidinger, E., Enbar, R., & Hassner, T. (2014). Age and gender estimation of unfiltered faces. IEEE Transactions on information forensics and security, 9(12), 2170–2179.
Foggia, P., Greco, A., Saggese, A., & Vento, M. (2023). Multi-Task Learning on the Edge for Effective Gender, Age, Ethnicity and Emotion Recognition. Engineering Applications of Artificial Intelligence, 118(1), Article ID 105651.
Guo, G., Mu, G., Fu, Y., Dyer, C., & Huang, T. (2009). A study on automatic age estimation using a large database. In IEEE 12th International Conference on Computer Vision (pp. 1986-1991).
Gupta, S. K., & Nain, N. (2023). Single attribute and multi attribute facial gender and age estimation. Multimedia Tools and Applications, 82(1), 1289-1311.
Han, H., Otto, C., & Jain, A. K. (2013, June). Age estimation from face images: Human vs. machine performance. In 2013 international conference on biometrics (ICB) (pp. 1-8). IEEE.
Hassan, O. M. S., Abdulazeez, A. M., & Tiryaki, V. M. (2018). "Deeply learned classifiers for age and gender predictions of unfiltered faces." The Scientific World Journal, 2020 (1289408).
Hu, Z., Wen, Y., Wang, J., et al. (2017). Facial age estimation with age difference. IEEE Transactions on Image Processing, 26(7), 3087–3097.
Jun, T. J., Eom, Y., Kim, D., Kim, C., Park, J. H., Nguyen, H. M., ... & Kim, D. (2021). TRk-CNN: transferable ranking-CNN for image classification of glaucoma, glaucoma suspect, and normal eyes. Expert Systems with Applications, 182, 115211.
Kendall, A., & Cipolla, R. (2017). Geometric loss functions for camera pose regression with deep learning. In Proc. - 30th IEEE Conf. Comput. Vis. Pattern Recognition, CVPR 2017 (pp. 6555–6564).
Lanitis, A. (2008). Comparative Evaluation of Automatic Age-Progression Methodologies. EURASIP Journal on Advances in Signal Processing, 2008(1), Article ID 239480.
Lehmann, E. L., & Casella, G. (1998). Theory of Point Estimation (2nd ed.). Springer.
Liu, N., Zhang, F., & Duan, F. (2020). Facial Age Estimation Using a Multi-Task Network Combining Classification and Regression. IEEE Access, 8(14), 92441–92451.
Lu, X., Ma, C., Ni, B., et al. (2018). Deep regression tracking with shrinkage loss. Proceedings of the European conference on computer vision (ECCV), 353-369.
Mohammed, A. I., & Tahir, A. A. (2020). A New Optimizer for Image Classification using Wide ResNet (WRN). Academic Journal of Nawroz University, 9(4), 1-13.
Pan, H., Han, H., Shan, S., & Chen, X. (2018). Mean-Variance Loss for Deep Age Estimation from a Face. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 5285-5294).
Ricanek Jr., K., & Tesafaye, T. (2006). MORPH: A longitudinal image Age-progression, of normal adult. In Proc. 7th Int. Conf. Autom. Face Gesture Recognit (pp. 0–4).
Rothe, R., Timofte, R., & L. Van Gool. (2015). DEX: Deep EXpectation of Apparent Age from a Single Image. In Proceedings of the IEEE international conference on computer vision workshops (pp. 10-15).
Shen, W., Guo, Y., Wang, Y., Zhao, K., Wang, B., & Yuille, A. L. (2018). Deep regression forests for age estimation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2304-2313).
Taheri, S., & Toygar, Ö. (2019). On the Use of Dag-CNN Architecture for Age Estimation with Multi-Stage Features Fusion. Neurocomputing, 329, 300–310.
Tahir, A. A. (2012). Integrating artificial neural network and classical methods for unsupervised classification of optical remote sensing data. EURASIP Journal on Advances in Signal Processing, 2012, 1-12.
Tingting, Y., Junqian, W., Lintai, W., & X.Yong. (2019). Three-stage network for age estimation. CAAI Transactions on Intelligence Technology, 4(2), 122-126.
Toshev, A., & Szegedy, C. (2014). DeepPose: Human Pose Estimation via Deep Neural Networks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1653-1660.
Xia, M., Zhang, X., Liu, W., Weng, L., & Xu. Y. (2020). Multi-Stage Feature Constraints Learning for Age Estimation. IEEE Transactions on Information Forensics and Security, 15, 2417-2428.
Yang, S., Lin, W., Lin, G., Jiang, Q., & Liu, Z. (2021). Progressive Self-Guided Loss for Salient Object Detection. IEEE Transactions on Image Processing, 30(4), 8426-8438.
Yang, T. Y., Huang, Y. H., Lin, Y. Y., Hsiu, P. C., & Chuang, Y. Y. (2018). SSR-NET: A compact soft stagewise regression network for age estimation. Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence (IJCAI-18), 1078–1084.
Zhang, C., Liu, S., Xu, X., & Zhu, C. (2019). C3AE: Exploring the limits of compact model for age estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 12587-12596).
Zhang, K., Gao, C., Guo, L., et al. (2017). Age Group and Gender Estimation in the Wild with Deep RoR Architecture. IEEE Access, 5, 22492–22503.
Zhang, Y., Liu, L., Li, C., & Loy, C. C. (2017). Quantifying Facial Age by Posterior of Age Comparisons. Proc. Brit. Mach. Vis. Conf. (BMVC), 126-145.
Zhang, Z., Song, Y., & Qi, H. (2017). Age Progression / Regression by Conditional Adversarial Autoencoder. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 5810–5818).