EE ZOOM Seminar: Tobamovirus Detection Using Deep Learning Algorithms on Hyper-spectral Images
https://tau-ac-il.zoom.us/j/87937383476?pwd=VcPMEYtX0bFrNmOgPgW007m1BwtJnY.1&from=addon
Meeting ID: 879 3738 3476
Passcode: 945671
Electrical Engineering Systems ZOOM Seminar
Speaker: Itai Friedman
M.Sc. student under the supervision of Prof. Noam Koengstein
Wednesday, 2nd July 2025, at 15:00
Tobamovirus Detection Using Deep Learning Algorithms on Hyper-spectral Images
Abstract
Tobamoviruses, such as Tomato brown rugose fruit virus (ToBRFV) and Cucumber green mottle mosaic virus (CGMMV), pose a significant threat to global agricultural productivity, particularly in high-value crops like tomatoes and cucumbers. These seed-transmitted viruses can spread rapidly, causing substantial economic losses and affecting food security worldwide. Early detection of these viruses in seeds is essential to prevent their spread and ensure healthy crop production. In this study, we propose a novel approach combining hyperspectral imaging (HSI) and deep learning techniques to detect Tobamovirus infections in tomato and cucumber seeds. A unique dataset of healthy and infected seeds was collected, utilizing custom-designed trays and a Visible and Near-Infrared (VNIR) camera for hyperspectral image acquisition. The goal of this research is to develop an accurate, non-invasive method for detecting infected seeds while exploring advanced deep learning architectures tailored for hyperspectral image classification.
To achieve this, we developed Hyperspectral Convolutional Vision Transformer (HCViT), a novel hybrid model that integrates components from Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), leveraging both local and global feature extraction capabilities. HCViT was evaluated using a held-out test set, achieving an accuracy of 0.94 for detecting infected tomato seeds, 0.78 for Ilan cucumber seeds, and 0.84 for Derby cucumber seeds. Comparative experiments demonstrated that HCViT outperformed standalone CNN and ViT models, highlighting the effectiveness of combining deep learning with HSI for early virus detection in seeds.
השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום בצ'ט של שם מלא + מספר ת.ז.
