Efficient detection of eyes on potato tubers using deep-learning for robotic high-throughput sampling

Abstract

Molecular-based detection of pathogens from potato tubers hold promise, but the initial sample extraction process is labor-intensive. Developing a robotic tuber sampling system, equipped with a fast and precise machine vision technique to identify optimal sampling locations on a potato tuber, offers a viable solution. However, detecting sampling locations such as eyes and stolon scar is challenging due to variability in their appearance, size, and shape, along with soil adhering to the tubers. In this study, we addressed these challenges by evaluating various deep-learning-based object detectors, encompassing You Look Only Once (YOLO) variants of YOLOv5, YOLOv6, YOLOv7, YOLOv8, YOLOv9, YOLOv10, and YOLO11, for detecting eyes and stolon scars across a range of diverse potato cultivars. A robust image dataset obtained from tubers of five potato cultivars (three russet skinned, a red skinned, and a purple skinned) was developed as a benchmark for detection of these sampling locations. The mean average precision at an intersection over union threshold of 0.5 (mAP@0.5) ranged from 0.832 and 0.854 with YOLOv5n to 0.903 and 0.914 with YOLOv10l. Among all the tested models, YOLOv10m showed the optimal trade-off between detection accuracy (mAP@0.5 of 0.911) and inference time (92 ms), along with satisfactory generalization performance when cross-validated among the cultivars used in this study. The model benchmarking and inferences of this study provide insights for advancing the development of a robotic potato tuber sampling device.

Keywords: FTA card; YOLO; machine vision; molecular diagnostics; potato pathogens; tissue sampling robot.

Figure 1

Potato tuber tissue sampling workflow (A) Desired sampling locations on a potato tuber indicated by four white arrows; (B) Manual sampling of potato tuber tissues onto FTA cards; (C) An FTA card containing samples from 25 potato tubers (four cores per tuber); (D) Nucleic acids from tuber tissue cores released onto the FTA card using a mechanical press; (E) FTA cards in envelopes ready to be shipped to a laboratory for downstream PCR-based pathogen detection.

Figure 2

A prototype of the robotic tuber sampling platform. The camera captures tuber images that are used for detection of eyes and stolon scar on the tuber. The sampling tool performs sampling on the identified instances and places the tuber cores on the FTA card.

Figure 3

Example images of potato tubers used to develop models for the detection of eyes and stolon scar. The yellow polygons denote the bounding boxes of the eyes. Red ovals denote regions that exhibit attributes similar to the eyes.

Figure 4

The image acquisition setup used for collecting potato tuber images dataset.

Figure 5

Workflow of the study outlining the development and evaluation process of identifying optimal object detection models for detecting eyes and stolon scars on potato tubers.

Figure 6

Examples of potato tuber images with the ground-truth (red boxes) and predicted bounding boxes (blue boxes) of the eyes and stolon scar.

Figure 7

The inference time (ms) (A) and number of parameters (millions) (B) versus mAP@0.5 for different YOLO object detectors for detecting eyes on potato tubers. The models belonging to the same YOLO detector are labeled with the same marker shapes and color.

Figure 8

Eye detection results of the YOLOv10m model across different confidence score thresholds (CST) on sample tuber images obtained from tubers of ‘Ciklamen’ (first row), ‘Purple Pelisse’ (second row), and russet cultivars (third row).