Greenhouse Legume Crop Recognition Dataset

In the agricultural field, with the development of smart agriculture, crop monitoring and management are facing more and more challenges, especially in the recognition of greenhouse crops. Current monitoring methods often rely on traditional manual observation, which is inefficient and prone to errors, unable to meet the needs of large-scale agricultural production. This dataset aims to provide high-quality legume crop recognition data to support research and applications of object detection technology. The dataset includes images of various legume crops, covering different growth stages and lighting conditions, capable of effectively improving the performance of recognition systems. Data collection was performed using high-resolution cameras in diverse natural environments to ensure data diversity and authenticity. Data annotation employs a combination of multiple reviews and expert evaluations to ensure accuracy and consistency of the annotations. The data storage format is JPEG, organized in a folder structure for easy subsequent processing and use. The dataset has excellent data quality, with annotation accuracy exceeding 95%, and consistency and completeness ensured through multiple reviews, solving the annotation error issues found in traditional methods. By introducing new data augmentation techniques, the dataset's performance in model training has improved by over 15%, greatly enhancing the practical application value of crop recognition systems.