Develop and deploy tools that use the geometric morphometrics concept to rapidly identify the invasive cotton jassid and similar leafhopper species from trap images, improving the ability to detect…

The U.S. Department of Agriculture Animal and Plant Health Inspection Service awarded West Virginia University Research Corporation a cooperative agreement worth $114,780 on June 1, 2026, to develop and deploy deep learning tools for rapid identification of the invasive cotton jassid and related leafhopper species from trap images.

The project team will develop, train, and validate transformer-based deep learning object detection models integrated with morphometric features to distinguish cotton jassid from look-alike native genera including Empoasca, Kyboasca, Jacobiasca, and Hebata. The team will build and beta-test a secure web-based platform enabling USDA and state cooperators to upload trap images for automated identification, with iterative performance improvement using user feedback and new data. Key deliverables include a curated, shareable collection of images and landmark datasets for geometric morphometrics; trained deep learning models capable of detecting and identifying cotton jassid and related leafhoppers; and documented, open-source code supporting future enhancement and deployment. The project targets at least 90% accuracy and precision, with performance improving as additional images accumulate. A beta-test web-based identification tool will provide rapid, standardized preliminary identifications to USDA and state cooperators, reducing identification backlogs and enhancing early detection capacity in domestic surveys by replacing slow, expert-dependent identifications with scalable image-based methods.

Performance occurs in West Virginia. The award runs through May 31, 2027, with project personnel including a principal investigator, postdoctoral computer vision engineer, and graduate research assistant conducting data collection, analysis, model development, and tool implementation.

Develop and deploy tools that use the geometric morphometrics concept to rapidly identify the invasive cotton jassid and similar leafhopper species from trap images, improving the ability to detect and respond to quarantine-significant pests in the united states and united states territories. This work will enhance early detection capacity in domestic surveys by replacing slow, expert-dependent identifications with an efficient, scalable, image-based method. Activities to be performed: the project team at west virginia university will develop, train, and validate deep learning object detection models (e.g., transformer-based architectures) integrated with morphometric features to distinguish cotton jassid from look-alike native genera such as empoasca, kyboasca, jacobiasca, and hebata. The project team will then build and beta-test a secure method for united states department of agriculture (USDA) and state cooperators to upload trap images for automated identification, iteratively improving performance using user feedback and new data. Project personnel, including the principal investigator, a postdoctoral computer vision engineer, and a graduate research assistant, will carry out data collection, analysis, model development, and tool implementation. Deliverables and expected outcomes: key deliverables include a curated, shareable collection of images and associated landmark datasets for GEO metric morphometrics, trained deep learning models capable of detecting and identifying cotton jassid and related leafhoppers, and documented, open-source code to support future enhancement and deployment. The project aims to achieve at least 90% accuracy and prec ision, with performance improving as more images are added over time. A beta-test stage, web-based identification tool will be made available to USDA and state cooperators for trap-image uploads, providing rapid, standardized preliminary identifications that reduce backlogs for expert taxonomists and diagnostic laboratories. Intended beneficiary(IES): primary beneficiaries include USDA animal and plant health inspection, plant protection and quarantine programs, state departments of agriculture, and other domestic survey programs that rely on sticky traps to monitor for invasive leafhoppers. These stakeholders will gain faster, more reliable identification of cotton jassid and similar species, enabling earlier detection, more timely response, and more efficient allocation of taxonomic and diagnostic resources. Longer-term, united states growers and the broader agricultural sector will benefit from reduced risk of undetected establishment and spread of cotton jassid and other quarantine-significant leafhoppers that can damage crops and vector plant pathogens.