MSc Seminar - Alexander Olpin

Title: Convolutional Neural Networks for Morphological Feature Identification in Pre-Harvest Crops  

Abstract:

Research in the field of Agritech focuses on applying technological solutions to solve traditional agricultural based problems, some of which include: crop yield prediction, weed vs crop classification and crop based feature extraction. In recent years Convolutional Neural Networks have experienced high levels of success with image classification tasks, but have yet to see wide spread application in the field of Agritech, likely due to their complex nature and need for large datasets. We propose the use of a Convolutional Neural Network for the purpose of identifying morphological features in pre-harvest agricultural crops. In order to test this approach we propose the development of a large simulated image data set to train and validate the Convolutional Neural Network model. A transfer learning approach will then be applied to examine how well the newly trained Convolutional Neural Network model generalizes to real world agricultural crop images.

Advisor: Dr. Deborah Stacey
Co-Advisor: Dr. Rozita Dara