Reliable potato leaf disease classification depends not only on classifier design but also on how disease-specific visual information is represented. Conventional deep learning classifiers mainly learn discriminative features from labelled images, while diffusion models in agricultural applications have largely been used for image synthesis and data augmentation. To address this gap, this study proposes a 1024-D Enhanced Class-Conditioned Latent Diffusion Model (EC-LDM) that employs latent diffusion directly for disease-aware representation learning and potato leaf disease classification. The framework integrates residual feature extraction, self-attention, and class conditioning with latent diffusion to learn structured disease-discriminative representations. Reconstruction, diffusion, classification, and KL-regularisation objectives are jointly optimised to refine the latent representation. Ablation studies compare frozen and fine-tuned 512-D configurations with the proposed 1024-D configuration to examine the effects of latent dimensionality and encoder optimisation. Experiments were conducted on the publicly available potato leaf disease (PLD) dataset, which contains early blight, late blight, and healthy classes. Five-fold stratified cross-validation was employed to evaluate the proposed model against eight Bayesian-optimised deep learning classifiers. Across five-fold stratified cross-validation, the proposed model achieved a mean accuracy of 98.69
Potato tuber diseases can adversely affect post-harvest quality, storage potential, and market value, creating a need for informative automated assessment. However, existing agricultural disease datasets are primarily designed for classification or detection and provide limited multimodal supervision for disease-oriented image captioning. We introduce two complementary contributions: (i) PotatoCapDS2026, a real-world multimodal dataset of 8016 potato tuber images covering nine health-condition classes, with structured symptom and severity attributes and five template-guided natural-language reference captions per image; and (ii) Disease-Aware Fusion BLIP (DAF-BLIP), a vision-language framework extending Bootstrapping Language–Image Pre-training (BLIP) by incorporating health-condition and attribute-aware representations with visual semantic features through a gated fusion mechanism for caption generation. DAF-BLIP is evaluated against convolutional neural network–long short-term memory (CNN+LSTM), Vision Transformer–long short-term memory (ViT+LSTM), BLIP-1, and BLIP-2 using BLEU, METEOR, ROUGE-L, CIDEr, and BERTScore. Across five repeated 80:10:10 holdout evaluations, DAF-BLIP achieved BLEU-4 of 0.7553, METEOR of 0.6993, ROUGE-L of 0.7417, CIDEr of 1.1801, and BERTScore-F1 of 0.9524. Statistical analysis identified significant differences in the evaluated performance measures, while expert evaluation provided additional assessment of caption quality and semantic relevance. The dataset and framework provide a resource for research on potato tuber health-condition image captioning and disease-oriented agricultural vision-language analysis.
Deficit irrigation (DI) is a sustainable approach to optimize water use while maintaining yield and quality of potato tubers, particularly in arid and semi-arid regions like the Mediterranean. This study assessed the impact of DI on the macro- (Ca, K, P, N, Mg) and micromineral (Cu, Fe, Mn, Na, Zn) element composition of four ‘early’ potato genotypes (G) across two growing seasons (S) in southern Italy. Three irrigation levels were tested: no irrigation (I0), 50
Single-date UAV vegetation indices are widely used to assess potato crop performance and estimate yield. However, whether their biological interpretation remains consistent across cultivars and growth stages is rarely examined. This study investigated how the relationships between UAV-derived multispectral and thermal canopy indicators and potato yield changed during crop development under contrasting irrigation in two cultivars with contrasting maturity classes (Anouk and Celandine). UAV surveys were conducted in July and August, and canopy metrics were related to final yield and tuber traits. The relationship between NDVI and yield differed markedly between cultivars and growth stages. In the early-maturing cultivar Anouk, July NDVI showed a strong positive correlation with yield (r = 0.87, p = 0.002), whereas August NDVI was not informative. In the late-maturing cultivar Celandine, the relationship reversed from a positive correlation in July (r = 0.56) to a strong negative correlation in August (r = −0.80, p = 0.009), indicating that greater late-season canopy greenness was associated with lower yield. Exploratory LOOCV Random Forest models achieved the highest yield prediction accuracy using July features for Anouk (R2 = 0.72) and August features for Celandine (R2 = 0.47), while multi-temporal features provided similar predictive performance. Greater temporal declines in NDVI were associated with higher yield in both cultivars, consistent with the change in canopy greenness acting as a phenological indicator of the transition from vegetative growth to tuber bulking. These results demonstrate that the interpretation of UAV-derived vegetation indices is strongly dependent on cultivar maturity and growth stage. A single UAV acquisition may therefore provide misleading assessments of crop performance when canopy phenology is not considered. Rather than serving as fixed proxies for yield, canopy vegetation indices should be interpreted within cultivar-specific and temporal contexts when used for potato monitoring and agronomic decision-making.
This study investigates the long-term structural transformation and supply responsiveness of potato production in Slovenia, a sector that has experienced a substantial and sustained contraction (more than 90
