PHIDIAS · PHIDIAS: Phenotyping with a High-throughput, Intelligent, Distributed, and Interactive Analysis System

In this proposal we present PHIDIAS, with the objective to provide an affordable, high-throughput, distributed, knowledge-sharing platform for phenotype collection and analysis. PHIDIAS will incorporate lessons learned from prior phenotyping efforts and will be a simple to use platform for acquiring (non-destructively and continuously) and analyzing phenotypes from plant experiments. Its novelty arises from the fact that it combines distributed affordable sensors (cameras) to collect images and transmit them over the Internet for centralized processing. With slight modifications to their software, commercial cameras can become powerful image acquisition devices. To increase the fidelity of our data, we acquire multiple sequential images of different parameters (focus, zoom, exposure) to generate high resolution, fully focused composite images, using sophisticated algorithms borrowed from computational photography running in processing servers. The same servers also process the images and extract relevant phenotyping information. All the data are stored in databases and offered to the user for exploration through a modern web-based Graphical User Interface. The users can even edit and view images online through the web. PHIDIAS learns from the user’s inputs, for training continuously the learning-based image processing algorithms, in order to increase performance. We will provide an arena of constant development and evolution by adopting open source, open development and affordable hardware standards. Thus, users can help PHIDIAS evolve by contributing data and functionality. As the knowledge in PHIDIAS’s database increases, and more biological databases are connected, complex meta-analysis studies can be performed bringing us closer to both forward and reverse hypothesis formulations. To the best of our knowledge there is no tool combining the above qualities, with the potential of providing an evolvable and sustainable phenotyping platform.