Hybrid AI Pipeline for Laboratory Detection of Internal Potato Defects Using 2D RGB Imaging

Abstract

The internal quality assessment of potato tubers is a crucial task in agro-industrial processing. Traditional methods struggle to detect internal defects such as hollow heart, internal bruises, and insect galleries using only surface features. We present a novel, fully modular hybrid AI architecture designed for defect detection using RGB images of potato slices, suitable for integration in industrial sorting lines. Our pipeline combines high-recall multi-threshold YOLO detection, contextual patch validation using ResNet, precise segmentation via the Segment Anything Model (SAM), and skin-contact analysis using VGG16 with a Random Forest classifier. Experimental results on a labeled dataset of over 6000 annotated instances show a recall above 90\% and precision near 100\% for most defect classes. The approach offers both robustness and interpretability, outperforming previous methods that rely on costly hyperspectral or MRI techniques. This system is scalable, explainable, and compatible with existing 2D imaging hardware.

Code and data availability

The paper's potato RGB image dataset (6000+ annotated instances) is explicitly not public due to laboratory confidentiality, with access only upon request; no public code, models, or data deposits are mentioned. All URLs in the text are cited references, not paper-specific assets.

No evidence-backed public reproduction asset is currently recorded.