In this study, we investigate the effectiveness of deep learning models with thermal images for gender categorization. In order to explore the possibilities of thermal imaging as a tool for gender identification, the study focuses on two sophisticated convolutional neural network (CNN) architectures: InceptionV3 and AlexNet. Thermal imaging is a powerful substitute for traditional visual data because it provides distinct physiological insights.A collection of thermal imaging datasets was assembled, methodically preprocessed, and divided into training and testing sets. For this comparison analysis, two well-known CNNs AlexNet, a fundamental model recognised for its straightforward yet efficient design, and InceptionV3, a complex model acclaimed for its inception modules were chosen. The training subset was used to carefully refine both models so they could accurately capture the subtleties of thermal-based gender traits.Accuracy was the main criterion used to assess the performance of the revised models on the testing subset. According to our results, InceptionV3 performs noticeably better than AlexNet, with an accuracy of 92.3% as opposed to 82.6% for AlexNet. This disparity in performance demonstrates how much better InceptionV3 is at identifying and deciphering minute thermal patterns and physiological indicators that are essential for precise gender categorization. This study highlights how sophisticated CNN architectures may improve gender categorization using thermal images, both in terms of accuracy and dependability. We provide a path for future research to investigate more intricate and integrated strategies, like multi-modal fusion and sophisticated feature extraction techniques, to further enhance the resilience of thermal-based gender classification systems by proving the efficacy of InceptionV3 over a more conventional model like AlexNet.