The work presented here addresses the complex decision processes in plant breeding by applying an artificial intelligence (AI) reasoning technique known as computational argumentation . Argumentation is a logic‐based formal methodology that analyzes evidence in support for and/or against particular conclusions. Argumentation is particularly useful for identifying and helping to resolve conflicts in an evidence base and underlying reasoning priorities. To date, argumentation has primarily been applied in a research context, for example, to analyze debates or legal questions, and has been linked to explainable AI to help users understand the outputs of an AI (e.g., machine learning) system. The contribution presented here incorporates a range of inputs (i.e., evidence) from professional and academic plant breeders to instantiate a set of logical arguments that support different, sometimes conflicting, plant breeding objectives. The use case demonstrated is the exemplar biofuel crop miscanthus. A prototype version of this system, Miscanthus‐AI , designed with a web interface for use by plant breeders, combines ranked lists of breeding priorities with explanations and visual references of genotypes. This article describes the key features of Miscanthus‐AI, the argumentation methodology and its relevance for plant breeding tasks, and the architecture of the developed system and resulting prototype, linking in‐field data (e.g., three‐dimensional reconstructions and high‐resolution images) to explain breeding traits.
An extensive, publicly available dataset is presented—the LAR Broccoli dataset—which contains 20,000 manually annotated images of broccoli heads captured from a moving tractor at an organic farm in the UK. The dataset contains images of the same row of broccoli heads recorded at 30 frames per second (fps) with three different cameras. Two off-the-shelf, relatively low-cost depth-sensing cameras were used, with the tractor moving at a speed of around 1 km/h, in addition to a webcam, with the tractor moving twice as fast. The utility of the dataset is demonstrated in four ways. First, three different state-of-the-art detector models were trained on the dataset, achieving an overall mean Average Precision (mAP) score of over 95% for the best-performing detector. The results validate the utility of the dataset for the standard task of in-field broccoli head recognition. Second, experiments with transfer learning were conducted, initialised with a smaller pre-trained broccoli detection model, and refined with the LAR Broccoli dataset. Third, we assessed the advantages of transfer learning not only using mAP but also according to time and space requirements for training models, which provides a proxy metric for energy efficiency, a practical consideration for real-world model training. Fourth, the cross-camera generalisation among the three camera systems was compared. The results highlight that testing and training detector models using different camera systems can lead to reduced performance, unless the training set also includes some images captured in the same manner as those in the test set.