Automatic Farmland Pest Monitoring Dataset

Current agriculture faces challenges in pest monitoring regarding real-time capability and accuracy. Traditional methods rely mostly on manual inspection, which is inefficient and prone to errors. Existing automated monitoring systems often fail to adapt to complex farmland environments, resulting in frequent occurrences of missed detections and false alarms. This dataset aims to provide high-quality data support for AI training by combining individual pests and feeding traces to improve the accuracy and real-time capability of pest monitoring. Data is collected using drones and ground cameras, covering different types of farmland and climatic conditions. We conducted multiple rounds of annotation and ensured data quality through consistency checks and expert reviews. Data is stored in JPG images and JSON format annotation information, which is clearly structured for subsequent processing and analysis. This dataset has a high annotation accuracy with annotation consistency exceeding 90% and completeness over 95%. We introduce new data augmentation techniques to enhance the model's generalization ability, which is expected to increase pest recognition rates by 15% and reduce false alarm rates by 30%, offering significant application value.