Early and accurate fire detection is essential for preventing and mitigating the severe consequences of fires and explosions, particularly in residential and indoor environments. Traditional detection methods, which often rely on sensors or manual monitoring, can be limited in responsiveness and reliability. To address these challenges, this study proposes an enhanced approach to fire and smoke detection based on the YOLOv11 object detection model combined with hyperparameter optimization. The model was trained and evaluated on the Home-Fire dataset, consisting of 6,500 annotated images of fire and smoke captured under diverse lighting conditions and tested under a variety of different fire environmental conditions. Experimental results show that YOLOv11 and its optimized version achieve high predictive performance, with mAP@0.5 of 0.934 and 0.940, respectively, and mAP@0.5-0.95 of 0.643 and 0.633, demonstrating clear improvements over previous YOLO versions. Moreover, with a model size of only 5.2 MB, it can be easily deployed on devices without requiring high hardware specifications, while consuming minimal system resources. The method is expected to significantly enhance the accuracy of fire and smoke detection, contributing to the advancement of AI-based solutions for public safety, smart home systems, and disaster prevention technologies.