Cauliflower Disease Identification and Classification: A Comprehensive Review

Abstract

Cauliflower (Brassica oleraeea var. botrytis) is a high-value cruciferous vegetable that is susceptible to bacterial, fungal, oomycete and viral diseases, which often have similar visual symptoms. This is an update of the 2024 survey and a critical review of the evidence on cauliflower published up to 11 August 2026, presented as a structured narrative review. Bibliographic records and publisher platforms were systematically examined, references from eligible studies were traced, and research on disease classification, lesion localization, severity estimation, postharvest defect recognition, and field deployment was assessed. The literature available shows a clear evolution from handcrafted descriptors and traditional classifiers to transfer learning, YOLO-based detection, explainable AI, federated and few-shot learning, lightweight networks, and multimodal RGB-depth analysis. The within-dataset accuracy of several transfer-learning studies was above 99%, while the mean average precision of the modified YOLOv8 and the RGB-plus-estimated-depth detector was 91.1% and 92.0%, respectively. While these results are promising, they are limited by the repeated use of the small VegNet data set, the use of augmentation to expand the data set, the lack of consistent validation procedures, the use of ambiguous disease labels, the lack of calibration analysis, and the lack of external testing. Immediate research priorities are identified as multilocation validation, plant-level data partitioning, pathogen-informed ground truth, uncertainty estimation, lesion-level explanation, and computationally efficient edge deployment. The human-health significance of this area is indirect, and more properly related to food availability, product quality, responsible pesticide use, and residue control than to unsupported assumptions that cauliflower diseases directly lead to human disease