30921 research outputs found
Sort by
Generative diffusion models for agricultural AI: plant image generation, indoor-to-outdoor translation, and human preference alignment
Get PDFThe success of modern agricultural artificial intelligence (AI) depends heavily on access to large-scale, diverse, and annotated plant image datasets. However, collecting such datasets in real-world field conditions is costly, labour-intensive, and constrained by seasonal and environmental variability. This thesis investigates the use of diffusion-based generative modelling to address these challenges through plant image synthesis, indoor-to-outdoor translation, and human preference-aligned fine-tuning. First, a Stable Diffusion model (SD-1.4) was fine-tuned with curated indoor and outdoor plant imagery to generate realistic, text-conditioned images of canola and soybean plants. Quantitative evaluation using Inception Score (IS) and Frechet Inception Distance (FID), along with downstream experiments on phenotype classification, demonstrated that synthetic images can effectively augment training data and improve model performance. Second, we explored image translation to bridge the gap between high-resolution indoor plant datasets and limited outdoor field imagery. By combining DreamBooth-based text inversion with image-guided diffusion, indoor plant structures were preserved while environmental contexts such as lighting, soil, and stress conditions were rendered according to outdoor semantics. Translated images were evaluated in a weed detection and classification task using YOLOv8, showing consistent improvements as synthetic data ratios increased. Finally, a preference-guided fine-tuning framework was developed to align generative outputs with expert judgements of quality and botanical realism. A reward model, trained on manually annotated scores, was integrated into a reward-weighted supervised fine-tuning procedure. The resulting preference-aligned model achieved higher subjective quality and stability, albeit with trade-offs in objective metrics such as FID. Overall, this work demonstrates that diffusion models can generate, translate, and refine plant images in ways that address data scarcity and domain gaps in agricultural AI. By coupling generative pipelines with expert feedback, this thesis introduces a pathway toward data-efficient and user-centred agricultural machine learning systems.I also received funding from NSERC.February 202
Summary of findings
No full textIn the fall of 2023 and winter of 2024, two online surveys were used to gain input on health care directives for vaccinations in Manitoba. In the first survey we asked questions about the general concept of health care directives and vaccinations, and proposed a few statements that could be used by individuals. We asked survey respondents to rate these statements and provide their specific comments and suggested revisions.
Based on the feedback from the first survey, we proposed one statement, and then asked for further input, suggestions, and comments. This report provides a general summary of the findings from both surveys
Understanding beam characteristics for the CERN WCTE experiment
No full textThe Water Cherenkov Test Experiment(WCTE) aims to understand the responses of the
detector for different charged particles such as π
±, p+, e±, µ±, K± and gamma rays with
momenta between 100 MeV/c to 1 GeV/c. It also aims to study the different physics
processes such as hadron scattering and pion scattering / absorption [1].
A beam test was conducted in July 2023 using the T9 beamline at CERN. This ex periment was carried out in a “low momentum configuration” and a “high momentum
configuration” or “tagged gamma configuration”. For the high-momentum configuration,
a magnet and a hodoscope were added to the detector setup. This thesis includes prepa ration of the tagged photon beam hodoscope and parts of the analysis of data collected
in 2023. In particular, the analysis studies will cover momentum calibration of detectors
like Time of Flight and Lead Glass Calorimeter, calibrating energy scales for lead glass
and finding its energy resolution are present. This thesis also includes estimating the
uncertainty on the momentum and energy scale in these detectors. This is a crucial
understanding required in both the tagged gamma setup and low-momentum setup.February 202
Elucidating factors that influence infection success and the behaviour of gastropod hosts of Echinostoma trivolvis lineage c
No full textWildlife trematodes are important to study due to their impact on wildlife populations, including the ability to modify the behaviour of their hosts. However, these parasites are challenging to study without a fully elucidated life cycle that can be completed in a laboratory. As such, a model for wildlife trematodes would be a great asset for learning about the ecology and behavioural manipulation found in this group. I chose Echinostoma trivolvis lineage c as a parasite model to study these topics in gastropod hosts. Chapter 1 determined how host species/morphotype, host size and miracidial dose affects infection success of gastropod first intermediate hosts. I confirmed reports from natural infections that Ladislavella elodes is a host along with it’s morphotype Stagnicola reflexa and further determined which size and exposure dose produced the highest infection success. I used this new information to obtain infected hosts for my second chapter of my thesis. Herein, I tested whether light and species influenced the navigational behaviour of potential second intermediates host L. elodes. I determined that snail velocity increased when exposed to a higher light level. Using these higher light conditions, I also found that there was no attraction to infected conspecific hosts by potential second intermediate hosts. Both experiments suggest that abiotic and biotic factors can alter snail navigational behaviour and may even influence whether host behavioural modification is observed. As such, laboratory-based studies of behavioural modification may be over or underestimating the strength and frequency of the behaviours that may be occurring in nature. Examples of naturally occurring behavioural modification may be missed entirely if not tested under the right context in the laboratory.Natural Sciences and Engineering Research Council of CanadaFebruary 202
Advanced data-driven methods for IOT sensor data and anomaly analysis in building environments
No full textIn commercial buildings, maintaining and controlling the indoor environment while balancing thermal comfort, building health, and energy consumption can be challenging. This thesis focuses on developing statistical and machine learning methods to study the indoor environment of commercial buildings using IoT sensors to identify issues that could affect building health or tenant comfort. Controlling the temperature within an acceptable thermal comfort range is crucial. To address this, we developed and applied statistical methods to identify locations with similar environmental conditions, which helps optimize the HVAC (Heating, Ventilation, and Air Conditioning) system. In this study, we introduced a novel method for identifying time series clusters using community detection in network analysis. Additionally, buildings may have significantly high-temperature ‘hot spots’ and low-temperature ‘cold spots’. To identify these locations, we developed an improved hot spot analysis approach. We also proposed a model to detect locations with abnormal behaviors due to environmental issues. Finally, we developed a Bayesian predictive model using Bayesian networks by introducing a novel method for learning network structures to enhance prediction accuracy and estimate the probability of encountering abnormal locations in the future.February 202
Use of newly selected Bacillus isolates to control enterotoxigenic Escherichia coli infections and improve gut health and performance of piglets
No full textTwo B. licheniformis isolates PF9 and HG76 were selected and subsequently characterized, and evaluated in the in vitro and in vivo studies. The isolate PF9 demonstrated tolerance to low pH and high bile salts and the ability to extend the lifespan of C. elegans infected with ETEC (P < 0.05). In the ETEC F4 challenged weaned piglet studies, villus height (VH) of jejunum and ileum in the pigs supplemented with B. licheniformis HG76 were higher than other pigs (P < 0.05). There was an increase in jejunal and ileal crypt depth (CD) in the HG76 group when compared to the non-challenged group (P < 0.05). The Vmax value of sucrase in the HG76 group was higher than the non-challenged group (P < 0.05). Piglets supplemented with B. licheniformis PF9 showed lower relative abundance of Bacteroidetes in the colon than piglets from the non-challenged group (P < 0.05). The supplementation of B. licheniformis PF9 increased the relative mRNA levels of tight junction proteins (P < 0.05).
Compared with the control group, dietary B. licheniformis HG76 significantly (P < 0.05) reduced the diarrhea index score. The ETEC F4 infection significantly induced the IPEC-J2 cell death, and this effect was significantly attenuated by B. licheniformis HG76 treatment (P < 0.05). The gene expression of ZO-1 was significantly increased during ETEC F4 infection with B. licheniformis HG76 treatment (P < 0.05). Moreover, the treatment of IPEC-J2 cells with B. licheniformis HG76 significantly alleviated the ETEC F4-induced elevation of the mRNA abundance of tumor necrosis factor alpha (TNF-α) (P < 0.05).
In conclusion, the supplementation of B. licheniformis PF9 relieved diarrhea, increased the relative mRNA level of several nutrient transporters, the gut barrier associated proteins, and digestive enzymes, and altered the microbiota composition in the ETEC F4 challenged weaned piglets. B. licheniformis HG76 could protect the intestinal cells from ETEC F4 infection by inhibiting inflammation and maintaining barrier integrity. The supplementation of B. licheniformis HG76 improved the intestinal morphology and increased digestive enzyme activity. Dietary supplementation with B. licheniformis PF9 and HG76 could relieve diarrhea severity after weaning.Agriculture and Agri-Food Canada, Manitoba Pork Council, Evonik Operations GmbH, and Canada Foundation for InnovationFebruary 202
Development of a rubber tire-derived asphalt binder product
Get PDFThe processing of rubber tire waste poses significant health and environmental risks, while potholes pose safety risks to commuters and patches often provide a temporary and unreliable solution to roadway maintenance. The incorporation of rubber tire waste into pothole patches offers a sustainable solution to both problems.
The use of rubber tire waste in asphalt pothole patches has shown improvement in structural properties. Improved structural properties include resistance to rutting, fatigue, thermal cracking, and high temperature failure. The main components of asphalt mixtures are aggregates and binder materials. Asphalt mixes have displayed success with modification of rubber for virgin components, such as aggregates and binder. Reliable Tire Recycling is a Manitoba company which converts rubber tire waste materials into value added products. This project explores the incorporation of crumb rubber obtained from Reliable Tire Recycling into an asphalt binder.
Four (4) asphalt binder samples were prepared with high-speed shear mixing which included a virgin binder sample as a control and samples with varying quantities of crumb rubber and warm mix additives. Quality of rubber was confirmed to meet technical specifications for moisture content, fiber and metal content, and crumb rubber size prior to mixing. Following mixing, each sample was subjected to verification tests to quantify success. The tests displayed results for the rheological and mechanical properties of the binders. The properties indicate the binders’ performance at high temperatures, resistance to common modes of failure, and workability.
The results displayed that an increase in rubber led to an increase in the viscoelastic properties. This indicates an improved resistance to rutting and fatigue failure. However, incorporation of warm mix additives increased viscosity more than double the binder mix at the same rubber quantity. This indicates poor workability of the mix with incorporations of warm mix additives.
Further research on rubber use in asphalt binders for pothole patching is recommended. Additional testing should be conducted on rubber-modified binder samples to analyze cold temperature performance and the effects of long-term aging. It is recommended to mix and test more formulas with varying quantities of crumb rubber to optimize the results
SPINEX-anomaly: similarity-based predictions with explainable neighbors exploration for anomaly and outlier detection
Get PDFAbstract This paper presents a novel anomaly and outlier detection algorithm from the SPINEX (Similarity-based Predictions with Explainable Neighbors Exploration) family. This algorithm leverages the concept of similarity and higher-order interactions across multiple subspaces to identify outliers. A comprehensive set of experiments was conducted to evaluate the performance of SPINEX. This algorithm was examined against 21 commonly used anomaly detection algorithms, and across 39 synthetic and real datasets from various domains and of a variety of dimensions and complexities. Furthermore, a complexity analysis was carried out to examine the complexity of the proposed algorithm. Our results demonstrate that SPINEX achieves superior performance, outperforms commonly used anomaly detection algorithms, and has moderate complexity (e.g., O(n log n × d)). More specifically, SPINEX was found to rank at the top of algorithms on the synthetic datasets and the 7th on the real datasets. Finally, a demonstration of the explainability capabilities of SPINEX, along with future research needs, is presented
Evaluation of physical and antioxidant properties of pigmented barley tortillas
Get PDFBarley contains bioactive compounds like β-glucan and polyphenols that promote health. Food processing, while necessary, can change barley's physical properties and adversely affect polyphenols' antioxidant activity through heat and the food matrix. Thus, identifying compounds post-production, including new chemical structures, is essential. Five barley genotypes (CDC Rattan, CDC McGwire, Roseland, HB21147, and HB21148) produced tortillas at varying pH levels (1, 3, 5, 7). Their texture and physical properties were assessed. Identification and quantification were achieved through high-performance liquid chromatography (HPLC) with standards. Assays on the acidified ethanol extract measured the total phenolic and anthocyanin content, along with antioxidant properties, which were evaluated via in vitro and ex vivo methods. CDC Rattan tortillas decreased in thickness significantly between pH 1 and 7. At pH 5, extensibility was lower, with the hardest texture across genotypes. HB21148 recorded the highest total phenolic (14.0±1.2 mg gallic acid equivalent/g) and total anthocyanin content (313.9±33.4 cyanidin-3-glucoside equivalent/kg) among raw forms. Catechin was most concentrated in HB21148, followed by HB21147. Roseland exhibited the highest trans-cinnamic acid (13.5±0.5 mg/kg). Gentisic acid appeared only at pH 5 in genotypes with high β-glucan, HB21147 and CDC Rattan. Antioxidant activities were similar at pH 3 and 7 per the FRAP assay. HB21147 and Roseland showed parallel activity across pH levels with the ABTS method. Caco-2 cell tests found CDC Rattan enriched with anthocyanins resulted in significantly higher CAA (cellular antioxidant activity) than HB21148 at 200 μg/ml. The study suggests pH 3 optimizes antioxidant properties with minimal texture impact, potentially through new compounds involving β-glucan and polyphenols. These insights may aid in developing food products that integrate natural bioactive compounds for health benefits.October 202
LL-37 and citrullinated-LL-37 modulate IL-17A/F-mediated responses and selectively suppress Lipocalin-2 in bronchial epithelial cells
Get PDFAbstract Background Levels of the human cationic antimicrobial host defence peptide LL-37 are enhanced in the lungs during neutrophilic airway inflammation. LL-37 drives Th17 differentiation, and Th17 cells produce IL-17A and IL-17F which form the biologically active heterodimer IL-17A/F. While IL-17 is a critical mediator of neutrophilic airway inflammation, LL-37 exhibits contradictory functions; LL-37 can both promote and mitigate neutrophil recruitment depending on the inflammatory milieu. The impact of LL-37 on IL-17-induced responses in the context of airway inflammation remains largely unknown. Therefore, we examined signaling intermediates and downstream responses mediated by the interplay of IL-17A/F and LL-37 in human bronchial epithelial cells (HBEC). As LL-37 can become citrullinated during airway inflammation, we also examined LL-37-mediated downstream responses compared to that with citrullinated LL-37 (citLL-37) in HBEC. Results Using an aptamer-based proteomics approach, we identified proteins that are altered in response to IL-17A/F in HBEC. Proteins enhanced in response to IL-17A/F were primarily neutrophil chemoattractants, including chemokines and proteins associated with neutrophil migration such as lipocalin-2 (LCN-2). We showed that selective depletion of LCN-2 mitigates neutrophil migration, functionally demonstrating LCN-2 as a critical neutrophil chemoattractant. We further demonstrated that LL-37 and citLL-37 selectively suppress IL-17A/F-induced LCN-2 abundance in HBEC. Mechanistic studies revealed that LL-37 and citLL-37 suppresses IL-17 A/F-mediated enhancement of C/EBPβ, a transcription factor required for LCN-2 production. In contrast, LL-37 and citLL-37 enhance the abundance of ribonuclease Regnase-1, which is a negative regulator of IL-17 and LCN-2 in HBEC. In an animal model of allergen-challenged airway inflammation with elevated IL-17A/F and neutrophil elastase in the lungs, we demonstrated that CRAMP (mouse orthologue of LL-37) negatively correlates with LCN-2. Conclusions Overall, our findings showed that LL-37 and citLL-37 can selectively suppress the abundance of IL-17A/F-mediated LCN-2, a protein that is critical for neutrophil migration in HBEC. These results suggest that LL-37, and its modified citrullinated form, have the potential to negatively regulate IL-17-mediated neutrophil migration during airway inflammation. To our knowledge, this is the first study to report that the immunomodulatory function of LL-37 enhances the RNA binding protein Regnase-1, suggesting that a post-transcriptional mechanism of action is mediated by the peptide