Thermal images are crucial for object detection in surveillance, security, industrial automation, and vehicular navigation due to their ability to capture heat signatures. However, complexities like low contrast, fluctuating thermal patterns, and lack of thermal-specific datasets necessitate the use of specialized algorithms for precise object identification and location. This research paper delves into the domain of object detection within thermal images, a crucial aspect of applications like surveillance and autonomous systems. Leveraging the YOLOv8 algorithm renowned for real-time and accurate object detection, this study focuses on enhancing detection performance in the challenging realm of thermal imagery. The methodology involves training on a meticulously curated dataset of 1898 thermal images, comprising 1518 for training and 380 for validation. The algorithm’s efficacy is demonstrated through a mean average precision (mAP50) of (82%) and map50–90 of (59.3%). These results validate YOLOv8 suitability for thermal imaging applications, propelling advancements in fields reliant on accurate object detection in thermally challenging conditions.