CornWeed Dataset: A dataset for training maize and weed object detectors for agricultural machines

There are many datasets available for training object detectors in non agricultural domains such as Autonomous Driving but these datasets fail to generalize well enough to an agricultural use-case. This dataset contains 3574 hand labelled images with bounding boxes provided in both YOLO and COCO dataset format. Additionally there are 4981 unlabelled images for testing purposes. All the images are hand labelled with bounding boxes with several labellers and reviewed. The dataset contains two class IDs namely maize and weeds. The dataset also contains a crop-row instance but has not been used in the accompanying paper but could be interesting for other future work. The images have been recorded in two resolutions i.e. 720 x 1280 and 480 x 640 covering two crop rows and single crop row respectively.