Integrating Morphological Diversity and Machine Learning-Based Phenotyping to Advance Melon (Cucumis melo L.) Classification and Management.

Melon (Cucumis melo L.) exhibits exceptional morphological, textural, and biochemical diversity, making infra‑specific classification and germplasm management particularly challenging. To develop a modern, trait‑driven framework, we evaluated 370 accessions from 50 countries using an extensive set of phenotypic descriptors covering vegetative, floral, and fruit characteristics. These included fruit shape and size, rind color and netting, flesh color and texture, soluble solids, seed morphology, and placental traits. Classical horticultural groupings showed strong correspondence with botanical–morphological analyses, confirming the continued relevance of traditional taxonomy while revealing inconsistencies in several broad or heterogeneous groups. Based on independent descriptors, we reconstructed a refined horticultural classification encompassing wild, feral, vegetable‑type, aromatic, and dessert melons. We defined several new Groups that capture previously under‑represented diversity, particularly within Eastern Europe and Central Asian germplasm. To enable scalable and objective characterization, we integrated machine‑learning phenotyping pipelines using image segmentation and morphometric trait extraction, allowing automated assignment of accessions to horticultural Groups. This approach provides reproducible, high‑throughput characterization suitable for large germplasm collections and breeding programs. Our results demonstrate that melon diversity can be resolved effectively through the integration of classical morphology with modern computational phenotyping, overcoming limitations of subjective scoring and enabling rapid, data‑driven classification. The unified, trait‑based system presented offers a robust foundation for systematic study, conservation, and utilization of melon diversity across its global range.