Abstract: Food security is a critical foundation for national stability and economic development, while pest and disease outbreaks in field crops pose a severe threat to grain production, necessitating efficient and precise monitoring and control measures. In recent years, deep learning-based pest and disease image recognition technologies have gained prominence, relying heavily on high-quality image datasets. However, the currently available public datasets face limitations in scale, coverage, and quality, hindering further breakthroughs in research and practical applications. This paper systematically reviews existing image datasets of major field crops, including rice, wheat, maize, and potato, focusing on their sources, key characteristics, and application scenarios. It also analyzes major challenges in data volume, diversity, and standardization. The findings reveal that while datasets exhibit a certain representativeness regarding collection time, location, and pest and disease types, improvements are needed in class balance, diversity, and cross-domain sharing. Summarizing and organizing these datasets provides technical support and theoretical insights to advance precision agriculture and ensure food security.

Key words: pest and disease images, field crops, datasets