Title: Age and gender prediction using Haar cascade algorithm and fine-tuned CNN framework

Authors: Danish Ali; Atta Ur Rahman; Bibi Saqia; Maqbool Khan

Addresses: Riphah Institute of System Engineering (RISE), Riphah International University Islamabad, 46000, Pakistan ' Riphah Institute of System Engineering (RISE), Riphah International University Islamabad, 46000, Pakistan ' Department of Computer Science, University of Science and Technology, Bannu, 28100, Pakistan ' Pak-Austria Fachhochschule: Institute of Applied Sciences and Technology, Haripur, Pakistan

Abstract: Age and gender prediction using unfiltered facial images is an essential task these days. We apply deep learning algorithm convolutional neural network (CNN) with multiple datasets such as UTKFace, Adience, and two custom datasets all combined for best performance. When we glanced over the existing works, we found that the existing models are efficient regarding age prediction, another problem was the computational power requirement, which is less useful in real-time and without a GPU. In this research, we used fewer layers to make this model usable in real-time and eliminate the requirement of GPUs for computational power. We employed the Haar cascade algorithm for face detection, and CNN with 12 layers for the other two modules, age prediction, and gender prediction. Combining different datasets and training each module on different balanced and unbiased datasets, we have achieved a higher accuracy regarding age prediction.

Keywords: age; gender; convolutional neural network; CNN; Haar cascade algorithm; UTKFace.

DOI: 10.1504/IJAISC.2025.148120

International Journal of Artificial Intelligence and Soft Computing, 2025 Vol.9 No.1, pp.22 - 34

Received: 13 Apr 2024
Accepted: 29 Oct 2024
Published online: 25 Aug 2025 *