Top view RGB image dataset of Urochloa hybrids for HTP and AI applications

This dataset comprises 2400 raw RGB images of Urochloa hybrids captured from top view perspective, providing a valuable resource for high-throughput phenotyping (HTP) and artificial intelligence (AI) applications. A subset of 255 images has been meticulously annotated with polygons to identify racemes, resulting in a total of 22340 annotations organized in the COCO format. The images were collected at the International Center for Tropical Agriculture in Colombia over the course of 2016 to 2017. With its extensive annotations and high-resolution imagery, this dataset is poised to advance research and development in plant phenotyping and computer vision, offering a unique tool for the study of tropical forage grasses.
Metodology:To collect the 2400 raw images, it was neccesary to do 12 field trips, wherein each session approximately 200 raw RGB images were captured, totaling 2400 images. This systematic approach was crafted to encompass the full spectrum of phenotypic diversity across various growth stages of the tropical forage grasses.