Age and Gender Classification from Retinal Fundus Using Deep Learning Tareq Obaid 1(B) , Samy S. Abu-Naser 1 , Mohanad S. S. Abumandil 2 , Ahmed Y. Mahmoud 1 , and Ahmed Ali Atieh Ali 3 1 Faculty of Engineering and IT, Alazhar University, Gaza, Palestine {tareq.obaid,abunaser,ahmed}@alazhar.edu.ps 2 Faculty of Hospitality, Tourism and Wellness, Universiti Malaysia Kelantan, Kota Bharu, Malaysia moha-nad.ssa@umk.edu.my 3 School of Technology and Logistics Management, Universiti Utara Malaysia UUM, Kedah 06010 Sintok, Malaysia Abstract. Since the rise of social media, age and gender classification have become increasingly important in a growing number of applications. Existing approaches, however, still have low accuracies when applied to real-world reti- nal fundus images. In the current study, we show that using Deep Learning (DL) to train representations can result in a considerable improvement in the age and gender classification performance. We present a DL model based on the pre- trained model named Xception. 26,000 retinal fundus images from the Kaggle library were collected for training the model suggested. The data was prepro- cessed before being divided into three parts (training, validation, and testing). The DL Xception model was assessed using the test data once it had been trained and cross-validated. The test results indicate that the ROC measure is 1.0, precision is 98.62%, recall is 98.62%, and f1-score is 98.61%, whilst accuracy is 98.62%. There is a prevalent non-awareness among clinicians regarding the changes in retinal variable variances among age and gender, emphasizing on the necessity of model explain ability of the age and gender classification of the images of retinal fundus. DL may assist clinicians to uncover new visions and illness biomarkers using the proposed method. Keywords: Gender · Age · Deep Learning · Classification · Retinal Fundus 1 Introduction DL is an Artificial Intelligence (AI) category which allows the autonomous learning and improvement of systems without depending on explicit programming. Machine learning entails the development of computer models which can autonomously access and utilize data. The learning starts by annotations or data like direct commands or experience that enable the identification of data patterns towards making better future decisions by the presented samples. The main objective is to teach a computer to absorb and perform like people, and eventually enable them to improve their learning on their own, by feeding those facts and information from observations and interactions in the real world [1, 2]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 F. Saeed et al. (Eds.): ICACIn 2022, LNDECT 179, pp. 171–180, 2023. https://doi.org/10.1007/978-3-031-36258-3_15