Sexual crimes are an increasing problem, with many cases difficult to identify due to the limitations of existing surveillance systems. This study aims to develop an Artificial Intelligence (AI)-based system using Convolutional Neural Network (CNN) and Support Vector Machine (SVM) for gender identification in order to support sexual crime investigations. The methods used include processing facial image datasets, training models using CNN for feature extraction, and SVM for gender classification. The results showed that the CNN model achieved an accuracy of 90.15%, while the SVM model only achieved an accuracy of 82.16%. Further evaluation with a confusion matrix showed that CNN was more accurate in classifying gender than SVM. With these results, the developed system has the potential to help authorities identify perpetrators of sexual crimes more quickly and accurately. The dataset used consists of 23,706 grayscale facial images of 48x48 pixels, with a balanced distribution of male and female samples. The CNN architecture includes three convolutional blocks and achieves 90.15% accuracy. Although designed for real-time operation, inference speed needs further validation using FPS or latency metrics on specific hardware platforms.