<p>Image-based potato disease research has achieved strong classification results, but many practical scenarios require a detector that can answer both what disease is present and where the relevant leaf evidence lies. This study presents PotatoScope, a compact detector derived from YOLOv8n for leaf-level potato foliar disease recognition. The method introduces a dual-axis morphology-contrast feature operator that preserves horizontal and vertical leaf cues while enhancing symptom-sensitive local texture, together with a contour-balanced Wasserstein regression objective that reduces the brittleness of overlap-only box learning on irregular biological shapes. To support detector-oriented evaluation, the classification-style potato leaf images are converted into a reproducible YOLO-format benchmark with whole-leaf bounding boxes, and a stock YOLOv8n baseline is trained on the same split for controlled comparison. Experiments on the public PlantVillage potato subset show that PotatoScope delivers stable high-accuracy detection across diseased and healthy categories while preserving real-time inference. The near-ceiling scores are interpreted in relation to the controlled single-leaf image conditions and should not be treated as lesion-level or field-deployment validation. The work provides a detector-oriented alternative to classification-only potato disease pipelines and offers a reproducible route for evaluating lightweight agricultural vision systems under clearly stated annotation assumptions.</p>