Deep learning for crop monitoring: a systematic review

A PRISMA-based systematic review of deep learning applied to crop monitoring. Eighty-seven studies were screened and synthesised across ten architecture families, with a structured comparison of reported performance, data modalities and evaluation practice.

Beyond cataloguing what works, the review maps what the field consistently fails to report: inconsistent validation protocols, limited geographic coverage outside high-income regions, and a gap between architectural novelty and demonstrated operational value. Those gaps directly motivated the controlled multi-modal experiment I am now running on Cameroonian crop data.

[TO COMPLETE: confirm the exact title, the full author list, the journal, volume, pages and DOI, then fill the fields in the sidebar.]