<p>Age and gender estimation are crucial in forensic odontology for identification and legal purposes. Conventional methods utilizing manual interpretation like Demirjian’s and Gustafson’s techniques are labor-intensive and prone to observer bias. Deep learning models, therefore, offer a promising alternative, enabling automated, accurate, and efficient assessments using dental radiographs. Therefore, the present study aims to predict age and gender using deep learning models and to compare their efficiency. Multiple convolutional neural network (CNN) architectures including ResNet18/50, DenseNet variants, EfficientNetB0, VGG16, MobileNetV3, and AlexNet were used; a total of 2341 panoramic radiographs were utilized. Images underwent preprocessing including normalization, resizing (to 224 × 224 pixels), and data augmentation (AutoAugment and RandAugment) to enhance model performance. Transfer learning with pretrained models—ResNet variants and DenseNet—was implemented with an ensemble strategy to optimize accuracy. Performance was assessed using metrics including accuracy, precision, recall, F1-score, and prediction time. DenseNet161 achieved the highest accuracy among individual models, with 90.4% accuracy for gender classification and 94% accuracy for age categorization. Ensemble models achieved accuracies of 94% for gender and 90.4% for age classification. While ensemble learning improved gender prediction, it did not outperform DenseNet161 in age estimation. Deep learning models provide a promising and objective approach for demographic estimation using panoramic radiographs. However, their clinical applicability requires further validation using larger, multi-center datasets and standardized evaluation protocols.</p>