Research Project: In Ground Root Segmentation using Deep/Transfer Learning

Location: Sugarcane Research

Project Number: 6052-13210-003-011-A
Project Type: Cooperative Agreement

Start Date: Aug 19, 2024
End Date: May 7, 2025

Objective:
T0 develop a method to analyze sugarcane root images. The cooperator will continue to analyze root images and develop the deep learning model, so that all minirhizotron images can be analyzed using the technology.

Approach:
In this project tens of thousands of photos have been taken of sugarcane roots from 44 access tubes from 0 to 100 cm below ground. The Cooperator and graduate student use deep and transfer learning to develop models to detect roots in the photos and extract the root pixels from the soil background to analyze the roots in another software called RhizoVision. They have already began this work and plan to continue until producing a complete product.