INTD 161

• Applications of AI in Plant Sciences was presented by R. Glen Uhrig from the Department of Biological Sciences, University of Alberta in April 2025.
• The Uhrig lab is situated on Treaty 6 territory, traditional lands of several indigenous nations and Métis people.

Uhrig Lab Members & Collaborators

• Sabine Scandola (Research Scientist AAFC)
• Guohui Lin (Prof. Computer Science)
• Nazmus Sakeef (MSc Student)
• Devang Mehta (Prof. KU Leuven)
• Curtis Kennedy (Undergraduate Reseacher/Research Associate)
• Broderick Wood & FoS SIP Interns

Photosynthesis

• Sunlight is essential for life.
• Sugar is the central product for plants.
• Photosynthesis Equation: 6CO2 + 6H2O → C6H12O6 + 6O2
• Plant products include: vitamins, rubber, wood, starch, fatty acids, carbohydrates, protein, and cellulose.

Capturing Light

• Plants use pigments and photoreceptors to capture light.
• Pigments include Chlorophyll a and Chlorophyll b.

Plant Interpretation of Light

• The circadian clock regulates 1/3 of all gene expression
• Covington et al., 2008 – Genome Biology, supports this point.
• Important in morning and evening loops.

Light Influence on Plant Development

• Multiple factors to consider include: increase of biomass (blue/red), petiole/stem elongation (far red), phytochemical composition & quantity (UV-B).
• Examples include Cannabis and Basil.
• Light modulation of plant growth and development can occur without genetic manipulation

LED Technology

• Light-Emitting Diode (LED) technology includes High Intensity Discharge, Fluorescent lighting and LEDs.
• High Intensity Discharge: Metal Halide (MH), High Pressure Sodium (HPS).
• Advantages of LED: Reduced heating, long lifetime, energy efficient, stable & homogeneous, precise control possible.

System of Study

• Arabidopsis thaliana (thale cress) is a model organism for plant studies.
• Life cycle of Arabidopsis thaliana: six weeks.
• First sequenced plant genome, with 11 iterations since 2000.
• Genetic resources for include TAIR, RIKEN, and NASC.
• It is a member of Brassicaceae with 3200 species. Including, kale, broccoli, cabbage, cauliflower, turnip, Chinese cabbage, common radish, and rapeseed.

Controlled Environment Chambers

• Environmentally Controlled Conviron Walk-in Chamber
• Perihelion LED Light Fixtures (G2V Optics)
• Top Mounted Automated Cameras
• PERIHELION VS908

Phenotype

• Phenotypic growth is evaluated using a combination of freely accessible and established PlantCV and ImageJ pipelines, it is also known as image analysis.
• Requires a mounted camera setup.
• Using a fully automated cloud acquisition system, plant growth can be monitored 24/7.

ML/DL in Plant Genotyping

• Model training is part of the application.
• Machine learning and Deep learning is used in plant genotyping
• Allows for differentiation between varying genotypes. For example, wild type (WT) vs. cry1/2.

ML/DL and Plant Genotyping

• Evaluation of results involves Precision, Recall, Accuracy, and F1-Score.

Light Use

• Hyperspectral cameras capture more than visible light, including diseases and deficiencies in crops.

Hyperspectral Imaging

• Measures light across a wide range of the electromagnetic spectrum.
• “Pseudo-Hyperspectral” opportunities include diffraction grating inside the camera and small aperture.

Pseudo-Hyperspectral Imaging

• Affordable option.
• Comes with parts list, CAD files, build instructions, etc.
• Light with known spectra is used to "fit" a line.
• RGB images have 3 values per pixel, and these hyperspectral images have ~250 values per pixel.

Application to Molecular World

• Using prediction, machine learning, big data, phenomics and genomics.
• Used in applications to molecular world, gene regulatory networks and biomarker development

Summary

• There are multiple opportunities in plants and agriculture.
• Computer vision combined with ML and DL applications.
• This includes horticultural Plant Growth and Field Phenotyping.
• There are extensive technology development opportunities!
• Enhanced Contextualization of Data, Gene Regulatory Networks, and Biomarker Development.

A Virtuous Cycle: Neuroscience and AI

• Neuroscience has inspired AI, and AI returns the favor.
• The presentation took place on April 3rd, 2025.
• The presenter was Dr. Quinn Lee from the University of Alberta.
• Dr. Lee is an Assistant Professor in the Department of Psychology and the Neuroscience and Mental Health Institute

What is Neuroscience?

• Neuroscience examines molecules, single neurons, pairs of neurons, and neural ensembles/circuits.
• It also studies brain systems that generate behavior, and ultimately, the behaving animal.

Key Figures in Neuroscience

• Santiago Ramón y Cajal (1852-1934)
• Camillo Golgi (1843-1926)
• Lorente de Nó (1902-1990)

Cajal vs. Golgi

• Cajal proposed the Neuron Doctrine: the nervous system comprises discrete, individual cells.
• Golgi proposed the Reticular Theory: the nervous system is a single, integrated network of cells.

The Rise of Computation

• Alfred Whitehead (1861-1947)
• Bertrand Russel (1872-1970)
• Walter Pitts (1923-1969)
• Warren McCulloch (1898-1969)

Early Neural Network Models

• McCulloch and Pitts created a logical calculus of ideas immanent in nervous activity in 1943.
• Their model's impact was non-existent in neuroscience.
• The McCulloch-Pitts paper and meetings inspired pioneers of computer science and engineering.

Key Figures in Early Computer Science

• John von Neumann (1903-1957)
• Frank Rosenblatt (1928-1971)

The Perceptron

• Rosenblatt developed the perceptron, which inspired work on information storage and organization in the brain.
• The perceptron illustrates fundamental properties of intelligent systems without being enmeshed in biological specifics.

Multilayer Perceptron (MLP)

• Multilayer perceptrons consist of Input layers, Hidden layers, and Output layers - Chan et al. (2023)

Understanding the Visual System

• Jerome Lettvin (1920-2011) worked with McCulloch and Pitts.
• The frog's eye informs the brain about the visual image through local pattern, independent of average illumination.
• Torsten Wiesel (1924-) and David Hubel (1926-2013) studied receptive fields of single neurons in the cat's striate cortex.
• Hubel and Wiesel (1962) researched neuronal response types

Visual System Deficits

• Individuals may experience vision issues due to traumatic brain injuries, developmental disorders.
• The visual system model implies that different aspects of visual experience are divided into brain areas.
• Types of deficits include visual agnosia and akinetopsia.

Principles of the Visual System

• The brain's visual system is organized in parallel, hierarchical systems.
• Complex representations are built from simple features like contrast and orientation.
• The system handles variations in size, shape, orientation, and location.
• This raises the question of building such principles into machines

The Neocognitron

• Kunihiko Fukushima (1936-) is known for the Neocognitron- Fukushima (1980)

Convolutional Neural Networks

• LeCun et al. (1989) worked on convolutional neural networks.
• Neural network kernels synthesized by the network function as feature detectors and are similar to vision systems - (Hubel and Wiesel 1962)” (LeCun et al., pp. 547).
• The question is prompted: do convolutional neural networks work like real brains?

ImageNet

• ImageNet Large Scale Visual Recognition Challenge (ILSVRC) has been underway from 2010 to present.
• It involves labels of over 14 million hand labeled images with over 20,000 image categories.

AlexNet

• AlexNet was created by Krizhevsky, Sutskever, and Hinton (2012)

Brain-Score

• Yammins & DiCarlo (2016) researched the area.
• Schrimpf et al. (2020) researched integrative benchmarking and models.
• Brain-Score is Integrative Benchmarks and Models at Scale.
• The Brain-Score platform can be applied to other domains of intelligence.
• There are implications of creating machines that are brain-like.

Open Problems in AI

• Major challenges remain in data and energy efficiency, navigation, complex actions/motor behavior, performing multiple tasks, and long-term memory.

Contact

• For questions, people can email [email protected].
• More information is on the Navigation and Memory Systems Lab website: https://sites.psych.ualberta.ca/NMSLab/