Greenhouse Image Classification Dataset

The current agricultural industry faces rapidly growing demands and the challenge of limited resources, especially in the area of greenhouse management. Existing datasets are often focused on single crops, lacking comprehensive classification for different types of greenhouses. This dataset aims to provide a rich image classification resource covering various types such as multi-span greenhouses, arch greenhouses, solar greenhouses, and daylight greenhouses to meet the needs of smart agriculture. Data collection is carried out using high-resolution cameras under natural light conditions to ensure image quality. For quality control, we employ multiple rounds of annotation and expert reviews to ensure label consistency and accuracy. The data is stored in JPG format and organized in a folder structure for ease of subsequent processing and training. The advantages of this dataset include its high annotation accuracy (95%), completeness (covering various greenhouse types), and the introduction of new data augmentation techniques to enhance model generalization capability, which is expected to improve recognition rates in crop monitoring tasks by at least 15%.