WeedID is an open source, searchable, weeds image platform designed to facilitate the development of machine learning algorithms for weed recognition in cropping systems. It brings together existing datasets with current collected images and enables users to contribute newly acquired weed imagery. The platform enables custom sorting and downloading of tailored datasets.

Large numbers of high quality, annotated weed images are essential for the development of weed recognition algorithms that are accurate and reliable in complex biological systems. Accurate weed recognition enables the use of site-specific weed control (SSWC) in agricultural systems, eliminating the need for wasteful whole field treatments. This approach substantially reduces weed control inputs and creates opportunities for the introduction of alternative weed control technologies that cannot feasibly be used as whole field treatments. SSWC relies on accurate detection (is a weed present) and identification (what is the species/further information on morphology) of weeds in agricultural systems (crop, pastures, rangelands and non-crop areas, etc.). Camera-based weed recognition using deep learning algorithms has emerged as a frontrunner in the development of accurate recognition algorithms suited to in-crop site-specific control. Training and development of algorithms requires significant quantities of high-quality, annotated images. WeedID is addressing this challenge by enabling the easy access and contribution of weed image data on an open source platform with search, dynamic filter and preview functions for custom dataset download capability.

WeedID is designed to be an opensource platform for sharing weeds images and annotations from around the world in every crop type! If you’re interested in contributing to the development and maintenance of the platform, the GitHub repository is available here. Alternatively, if you have images to upload then contribute those too. If you’re interested in contributing annotated images to WeedID, the general process across the data collection, annotation and finally upload pipeline is outlined below. Weed image collection