DeepRoot: Automated Root Analysis for Minirhizotrons

The importance of fine roots to the whole root and the whole plant lifespans is well recognized, but the fine roots dynamics and their response to multiple environmental factors are subject to research. Such research is important to understanding and mitigating the effects of climate change and can also provide a path to grow healthier forests, plantations, crops and commercially sold plants. Minirhizotrons offer a means to facilitate such research, but the state of the software for their image analysis still implies mostly manual inspection. Such interaction with the data is extremely laborious, tedious and error prone. Tech-X will address this problem by developing a software application for automated processing of images from minirhizotrons. This software will automatically detect fine roots and quantify their traits such as length, density, diameter, color, branching order, and nodulation and track these traits in time. This tool will be able to analyze images in multiple formats and perform efficient image analysis with the minimal human intervention. This software will take advantage of the recent developments in Machine Learning, provide an intuitive cross-platform Graphical User Interface and GPU implementation for speed. In Phase I we will prototype novel Machine Learning algorithms for traits identification in Python. We will investigate applicability, flexibility and accuracy of the developed tools on a variety of selected minirhizotron images. The proposed software will be available to multiple researchers in the national laboratories and universities performing studies on climate change. Commercial applications include farms growing Christmas trees, crops, vineyards, oil palm plantations, rubber tree plantations and fruit plantations.