USU-Corn-WeedDB: A UAV RGB Image Dataset for Multi-Species Weed Detection in Forage Corn
Researchers have introduced USU-Corn-WeedDB, a new dataset designed to improve weed detection in forage corn using drone imagery and deep learning. The dataset, collected from a commercial field in Utah, contains 8,800 image patches, with 800 manually annotated for three common weed species. This resource aims to address the scarcity of field-representative training data, which has limited the development of site-specific weed management systems. Initial tests with various object detection models showed competitive performance, indicating the dataset's utility for developing efficient AI-powered agricultural tools. AI
IMPACT Enables development of more accurate AI models for precision agriculture, potentially reducing crop loss and herbicide use.