Manual Harvesting Action Detection Dataset

The current agricultural sector faces challenges of labor shortages and low production efficiency, especially in the harvesting stage where traditional manual picking methods can no longer meet the growing market demand. Existing automated picking solutions often suffer from losses due to inaccurate operations, lacking efficient action detection technology. This dataset aims to solve the problem of accurately identifying and detecting manual harvesting actions to enhance the automation level and efficiency of agricultural production. High-definition capture equipment is used during the dataset construction process, with image capturing under natural light environments to ensure diversity and authenticity of the data. To ensure data quality, multiple rounds of annotation, consistency checks, and expert reviews are conducted as quality control measures, with the final data stored in JPEG format and organized into a hierarchical file structure to facilitate subsequent applications and analysis.