Full TGIF Record # 333524
Item 1 of 1
Web URL(s):https://scisoc.confex.com/scisoc/2023am/meetingapp.cgi/Paper/153110
    Last checked: 12/06/2023
Publication Type:
Content Type:Abstract or Summary only
Author(s):Shahoveisi, Fereshteh; Waldo, Benjamin
Author Affiliation:Shahoveisi: Presenting Author and Ag Solutions, College Park, MD; Waldo: United States Department of Agriculture, Beltsville, MD
Title:Advancing nematode identification in turfgrass using image processing and deep learning
Section:Turf pest managment oral II
Other records with the "Turf pest managment oral II" Section

C05 turfgrass science
Other records with the "C05 turfgrass science" Section
Meeting Info.:St. Louis, Missouri: October 29-November 1, 2023
Source:ASA, CSSA, SSSA International Annual Meeting. 2023, p. 153110.
Publishing Information:[Madison, Wisconsin]: [American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America]
# of Pages:1
Abstract/Contents:"Nematodes present significant challenges to the management of turfgrass, resulting in adverse effects on quality, density, and playability. Timely and precise identification of plant-parasitic nematodes is important for effective management practices and minimizing excessive input costs. Leveraging image processing and deep learning techniques can greatly enhance our ability to identify and distinguish these nematodes. This study aims to explore the potential of deep learning algorithms in the accurate identification of plant parasitic nematodes that primarily affect turfgrass. A dataset consisting of 23,813 nematode images belonging to seven groups, including lance, lesion, ring, root-knot, stunt, spiral, and cyst nematodes, was employed to train and validate four different models including EfficientNet B0, EfficientNet B4, MobileNetV2, and VGG-16. Adam Optimizer has been reported as the most efficient optimizer in similar studies. Hence, we used the Adam optimizer and tested different learning rates. Additionally, we assessed different validation splits, including 20-80%, 30-70%, and 40-60%, and various dropout thresholds. The number of epochs varied between 100 and 200, depending on the models. EfficientNet B0 outperformed other models in the preliminary analysis where it resulted in an average accuracy of over 95% across training and validation sets with a learning rate of 0.000005 and a validation split of 20-80% (test-train, respectively). The average training and validation loss values of 0.13 and 0.18, respectively, indicated that the model was not overfitted. These findings demonstrate the potential of employing deep learning algorithms to develop applications that can be utilized in diagnostic and research laboratories, enabling accurate and fast identification of nematodes. Such tools would greatly benefit laboratories with limited personnel trained in nematode identification, enabling them to accurately identify and classify plant parasitic nematodes."
This item is an abstract only!
ASA/CSSA/SSSA Citation (Crop Science-Like - may be incomplete):
Shahoveisi, F., and B. Waldo. 2023. Advancing nematode identification in turfgrass using image processing and deep learning. Agron. Abr. p. 153110.
Fastlink to access this record outside TGIF: https://tic.msu.edu/tgif/flink?recno=333524
If there are problems with this record, send us feedback about record 333524.
Choices for finding the above item:
Web URL(s):
    Last checked: 12/06/2023
Request through your local library's inter-library loan service (bring or send a copy of this TGIF record)