Vision Normalizing Flows for the probability-informed detection of banana diseases from in-field images

Banana diseases impose severe production losses in tropical smallholder farming systems, yet accurate in-field visual diagnosis remains difficult: symptom expression varies across cultivars and growth stages, and several diseases produce morphologically overlapping foliar signs. We developed a probabilistic image-recognition framework for detecting five economically important banana diseases — Xanthomonas Wilt, Banana Bunchy Top Disease, Fusarium Wilt (Panama disease), Yellow Sigatoka, and Black Sigatoka - from in-field photographs, without any disease-specific fine-tuning of the vision backbone. The approach extracts frozen 1,152-dimensional embeddings from the DINOv3 vision foundation model and couples them with a conditional normalizing flow, trained on four publicly available datasets comprising approximately 95,000 images across eight classes spanning diseased banana plants, healthy tissue, non-banana vegetation, and general natural imagery. On an independent test set the model achieved F1 scores exceeding 0.98, average precision values of 0.9678–0.9998, and AUROC values of 0.9967–0.9998 across all five diseases evaluated as binary detection problems. Multi-class accuracy was near-perfect, with limited confusion between Yellow Sigatoka and Black Sigatoka — a biologically plausible ambiguity attributable to overlapping early-infection foliar symptoms. Because the normalizing flow estimates explicit conditional probability densities rather than decision boundaries, two complementary log-likelihood ratios can be derived: a disease ratio comparing each disease class against healthy banana, and a plant ratio comparing banana against non-banana imagery. Together these define an interpretable two-dimensional diagnostic space that simultaneously quantifies evidence for disease presence and image relevance, cleanly separating diseased plants, healthy plants, and out-of-distribution images while flagging uncertain predictions for confirmatory testing. Batched inference costs 6.4 ms per image on a single TPU core, supporting deployment at surveillance-programme scale, providing a scalable, uncertainty-aware diagnostic tool for smallholder farming systems.