This paper provides an in-depth review of fire and smoke datasets collected over the past two decades. Despite advances in artificial intelligence (AI) and computer vision (CV) technologies, good quality data is essential for developing effective fire management systems. This paper analyzes the characteristics of various fire and smoke datasets (type, size, format, collection method, geographical diversity, imaging modality (RGB, thermal, infrared), etc.), summarizes the strengths and weaknesses of each dataset, and presents experimental analysis results using state-of-the-art algorithms such as ResNet-50, DeepLab-V3, and YoloV8. This was done with the aim of contributing to research and technological advancement for fire management.