Plant Phenotyping and Imaging Research Centre

The Plant Phenotyping and Imaging Research Centre (PIRC) is a new $37M initiative at the University of Saskatchewan that will drive transformative innovation in plant breeding through powerful new computational techniques to digitize desired crop traits (phenotypes) and link them to specific genes. The project involves exciting research across many areas of Computer Science including: image processing, machine learning, bioinformatics, high-performance computing, software engineering, programming languages, data analytics, computer graphics and human-computer interaction. PIRC builds on the University of Saskatchewan's internationally-renowned food security and computer science strengths and will accelerate crop development, bolstering Canada’s agricultural leadership and improving global food security.

Graduate Student Opportunities Available

We are searching for bright and enthusiastic individuals to join our team and make big data analysis in agriculture a reality. The ideal candidate will have strong computer programming skills and a keen interest in computational biology and computational agriculture research. Prior experience with machine learning, image processing, bioinformatics, digital signal processing, pattern recognition, software engineering, robotics, and/or human-computer-interaction is also desirable.

You will join a dynamic, interdisciplinary team focused on developing next generation digital agriculture and supporting the world-wide community of plant breeders. You will also have the opportunity to interact with collaborators in imaging science at the Canadian Light Source synchrotron and in plant biology at the Global Institute for Food Security at the University of Saskatchewan.

M.Sc. students will be funded at the rate of $20,000/yr for two years. Ph.D. students will receive $23,000/yr for at least 3 years. Eligibility for scholarships will require that the student have a GPA of at least 80% at the time of admission and maintain a GPA of at least 75% for the duration of their funding period.

Data Acquisition and Analysis (DANA)

What features should be extracted from images and sensor data to determine a digital phenotype?

Researchers

Ian Stavness, Lead
Kevin Stanley, Co-lead
Mark Eramian
Mike Horsch
Eric Neufeld
The primary purpose of DAnA is to provide succinct metrics of phenotype traits of species of interest and automated methods of extracting that information from source images. The secondary goal is to provide infrastructure, software and acquisition support for the capture of field-level multispectral drone imaging and point environmental data.

Data Acquisition Research       

Primarily staff charged with maintenance, deployment, acquisition and data cleaning of field data.

Masters level research on applying known acquisition algorithms to this context

  • Optimization
  • Demonstration
  • Extention

Data Analysis Research

Primarily PhD students
  • Applied research in difficult area for segmentation
  • Interesting collaborative research in feature selection
  • Data fusion research for metrics
Research areas involved
  • Computer vision
  • Machine learning
  • Data fusion

Skills

  • Sensor network and drone data acquisition
  • Data cleaning, provenance and database maintenance
  • High performance computing

Competencies and Cognates

  • Image processing
  • Sensor Fusion
  • Machine Learning

Bioinformatics for Linking Genotypes to Phenotypes

What is the optimal link between digital phenotypes and genotypes?

Researchers

Tony Kusalik, Lead
Michael Horsch, Co-lead
Matthew Links, Co-lead
Use machine learning to (organize and analyse plant data in order to) generate possible connections between phenotype and genotype

Determine whether co-occurrence (e.g. phenotype x with genotype y) is powerful enough to provide actionable insight to plant breeding

Research areas involved
  • Machine Learning
    • Bioinformatics
    • Statistics
    • Plant breeding
  • Plant molecular biology
  • Plant genetics
  • Plant genomics

Big Data Analytics

How do we efficiently store, process and analyze huge amounts of phenotypic and genotypic data?

Researchers

Kevin Schneider, Lead
Chanchal Roy, Co-lead
Dwight Makaroff
Kevin Stanley
Nadeem Jamali
Chris Dutchyn
Mark Eramian
Regan Mandryk
Derek Eager

What programming abstractions, data management and processing techniques can be used with huge, heterogeneous collections of phenotype and genotype data, so as to support a wide range of types of queries varying in their delay, cost, and quality-of-answer requirements?

Cluster-based computing using data analytics software stacks supporting both disk-based and in-memory processing can, with extension and optimization, provide a suitable basis for this application domain.

How can we best represent, integrate and store diverse, multi-dimensional digital crop phenotype data so that it is conducive to collection, augmentation and analysis?

Research areas involved

  • Distributed High Performance Computing
  • Software Research (Design, Engineering & Programming Languages)
  • Visualization and Analytics

Web Systems and Collaboration

How do we support the global plant breeding community in sharing and exploiting phenotype-genotype data?

Researchers

Carl Gutwin, Lead
Ralph Deters, Co-lead

How to enable collaboration between researchers, breeders, and public.

Applications that solve specific problems

  • General types of activities our systems can support:
  • Searching (visual, meta-data, contextual)
  • Visualization (of data, analysis results, processes, and interaction history)
  • Communication (between researchers, or to the public)
  • Coordination (of collaborative work efforts)
  • Interfaces for analysis and simulation (e.g., front ends for specifying the parameters of analysis)
  • Data-gathering (e.g., crowd-sourced image acquisition)
  • Annotation (commenting or documenting analyses)
  • Education (of the public, of researchers)

Research areas involved

  • Use of RESTful web services
  • Support for edge-computing

Algorithmic Modeling of Digital Plants

How can we model plant phenotypes using algorithms in ways that are useful for breeding?

How can we model plant phenotypes using algorithms in ways that are useful for breeding?

Researchers

Przemyslaw Prusinkiewicz, Lead
Ian McQuillan, Co-lead
Mark Keil
Mik Cieslak
Pascal Ferraro
What are the mechanisms of plant development (genes and molecular processes to phenotypes)?

Are there universal principles?

How do the answers apply to breeding?

Research areas involved

  • Modeling methods (e.g. L-systems) and software (vlab)
    • Simulation programs
    • Tools for exploring and analyzing models
    • Model management (storage, retrieval, sharing)
  • Construction of specific models
  • Understanding of fundamental processes
    • Mechanisms of genotype to phenotype mapping
  • Mathematical methods and fundamental questions
    • Properties of self-organization
    • Competition for and colonization of space

If you would like more information about the Plant Phenotyping and Imaging Research Centre (PIRC) at the University of Saskatchewan, please contact our Computer Science Graduate Program Assistant.