Multispectral fine-grained classification of blackgrass in wheat and barley crops

Madeleine Darbyshire, Shaun Coutts, Eleanor Hammond, Fazilet Gokbudak, A. Cengiz Öztireli, Petra Bosilj, Junfeng Gao, Elizabeth Sklar, Simon Parsons

Abstract: Highlights•A novel dataset of 15,000+ multispectral images to detect blackgrass in cereal crops.•Benchmarks using ResNet-50, EfficientNet B4 and Swin B on images from unseen fields.•NIR is highly effective for blackgrass recognition.•Performance improves with more data, up to 6,000 images.