High-throughput greenhouse imaging for phenotyping sunflower basal stalk rot

Authors: ANGIDI, SRUSHTIDEEP - North Dakota State University; DAS, ANUP - North Dakota State University; PASCHE, JULIE - North Dakota State University; FLORES, PAULO - North Dakota State University; Underwood, William

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 8/1/2026
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Sclerotinia basal stalk rot (BSR), caused by Sclerotinia sclerotiorum, is a major disease of sunflower, and greenhouse screening for resistance often relies on repeated visual assessments, which are labor-intensive and subject to observer bias. This study developed an automated image-based phenotyping approach for detecting BSR symptoms in sunflower. Red-green-blue (RGB) images were collected from greenhouse-grown sunflower plants root-zone inoculated with S. sclerotiorum isolate NEB274 using an in-house high-throughput imaging platform fitted with a Canon EOS Rebel T7i camera. Images captured disease progression, including wilting, senescence, and plant mortality. A total of 801 images representing different stages of symptom development were used for training and validation, and 60 independent images were used for testing. Eight You Only Look Once (YOLO)-based convolutional neural network (CNN) object detection models were evaluated, and YOLO12n showed the best performance, with a validation mean average precision at 50% (mAP@50) of 0.871 and a test mAP@50 of 0.773. The selected model is currently being further evaluated across additional image sets to assess its robustness under greenhouse conditions. This approach demonstrates the potential of integrating high-throughput imaging and deep learning to improve the efficiency, scalability, and consistency of BSR phenotyping and to support resistance screening in the sunflower breeding program.