
Accurate nitrogen status assessment in maize is critical for optimizing fertilizer application, yet smallholder farmers lack accessible diagnostic tools. Conventional methods are time-consuming, destructive, and require technical expertise, leading to suboptimal nitrogen management. This study aimed to develop and evaluate deep learning models for classifying maize nitrogen status across five fertilizer levels (0, 30, 60, 90, and 130 kg N ha⁻¹) and assess their applicability under smallholder farming conditions. Field and pot experiments were conducted at Sokoine University of Agriculture, Tanzania, using randomized complete block design with four replications. A total of 3,290 leaf images were collected between 31-59 days after planting to capture nitrogen deficiency symptoms following urea top-dressing at 14 days.