New Datasets on Artificial Intelligence for Weed Identification

Abstract

This paper presents the data sets acquired in the context of the ROSE agricultural robotics challenge. This database represents crop plants (maize and bean) associated with different types of weeds in outdoor field conditions with important agropedoclimatic disparities depending on the years of acquisition and the sensor architectures used to build them. These databases have been designed to support the development of agricultural robotics, with a focus on intra-row weeding.