Lakehead University Knowledge Commons

Knowledge Commons is an open access repository for scholarship and research produced at Lakehead University. It is a free and secure repository for LU faculty, students, staff, and researchers to preserve and present their scholarship.

Communities in KnowledgeCommons

Select a community to browse its collections.

Now showing 1 - 3 of 3

Recent Submissions

Wind load evaluation of modular housing structures for Canadian Indigenous regions
(2026) Brown, Tristen; Elshaer, Ahmed; Issa, Anas; Wang, Jin; Liu, Kefu; El-Gendy, Mohammed
Indigenous communities across Canada face a persistent housing crisis characterized by housing shortages, overcrowding, dwellings requiring major repairs, and housing costs that exceed 30% of household income. These challenges are exacerbated in remote and northern regions, where communities experience pronounced climate change impacts, including stronger wind events, colder temperatures, and increased environmental uncertainty. Geographic isolation, limited access to construction resources, and harsh climate conditions make it difficult to construct, maintain, and sustain conventional housing throughout its service life. Accordingly, resilient and cost-effective housing solutions suited to these environments are urgently required. This thesis investigates modularly constructed (MC) housing as a sustainable solution for Indigenous communities in northern Canada. MC structures offer advantages in constructability, transportation efficiency, and adaptability to remote conditions while providing opportunities to improve durability and reduce material waste. Although interest in modular systems is growing, previous research has focused primarily on mid- and high-rise structures, with comparatively little attention paid to low-rise MC configurations. Research addressing the combined challenges of extreme wind loading and limited climate data in northern regions is also scarce. To overcome these limitations, this research develops an integrated framework that combines climate data reliability assessment, advanced extreme wind prediction methodologies, and structural performance evaluation of the MC configurations. The first objective (Chapter 2) of this research is to review the ongoing challenges that Indigenous communities throughout Canada face in housing. It explores the structural and socio-environmental challenges affecting Indigenous housing, including climate impact, material deterioration, and constructability limitations. This study proposes modular housing as a sustainable and adaptable solution. Next (Chapter 3), the thesis examines the spatial relationship between Indigenous communities and nearby weather stations to assess data availability and reliability. The analysis identifies significant gaps in meteorological coverage and demonstrates that limited record lengths and spatial variability can substantially influence extreme wind predictions. Third, Chapter 4 applies the Up-Crossing Rate (UCR) analysis to estimate extreme wind velocities using limited wind datasets. Unlike traditional extreme value analysis methods, which require long-term records, the UCR approach utilizes the full wind speed time history to provide reliable predictions for shorter datasets. The method is validated against established statistical approaches, showing strong agreement for long-term return periods while maintaining applicability in data-scarce environments. The fourth objective (Chapter 5) introduces a record length coefficient to correct for the systematic bias associated with short-duration datasets. Using data from over 200 meteorological stations across Canada, the study quantifies the influence of record length on extreme wind predictions and develops a record length coefficient to account for limited datasets. This approach improves the reliability of predictions derived from limited data and supports their use in engineering designs. The final objective of this thesis (Chapter 6) is to investigate the aerodynamic performance of modular housing configurations evaluated using computational fluid dynamics (CFD). Experimentally validated simulations were used to analyze single- and multi-module low-rise structures subjected to wind loading. The results indicate that modular configurations, particularly those involving stacking and overhangs, can generate significantly higher localized pressure coefficients than those predicted by current design codes. Comparisons with the National Building Code of Canada reveal potential underestimation of critical wind loads.
Evaluating architecture scalability and transfer learning in urban scene segmentation using explainable AI
(MDPI, 2026-03-01) Hatkar, Tanmay Sunil; Pandey, Abhinav; Ahmed, Saad Bin
Semantic segmentation plays a pivotal role in autonomous driving, enabling pixel-level understanding of road scenes. Although transformer-based models such as SegFormer have shown exceptional performance on large datasets, their generalization to smaller and geographically diverse datasets remains underexplored. In this work, we analyze the scalability and transferability of SegFormer variants (B3, B4, B5) using CamVid as the base dataset. We perform cross-dataset transfer learning to KITTI and IDD, evaluate class-level performance, and explore explainable AI via confidence heatmaps. Our findings show that SegFormer-B5 achieves the highest accuracy (82.4% mIoU) on CamVid, while transfer learning from CamVid improves mIoU on KITTI by 2.57% and enhances class-specific predictions in IDD by over 70%. These results highlight the practical potential of SegFormer in real-world segmentation systems and the interpretability benefits of confidence-based visual analysis.
QuantFormer: A hybrid quantum classical transformer for hyperspectral image classification
(PMLR, 2026) Lunia, Jay Vinit; Ahmed, Saad Bin
Hyperspectral image (HSI) classification is challenging because each pixel has hundreds of spectral bands while only a small number of labelled samples are available. This paper presents QuantFormer, a hybrid quantum–classical transformer that embeds a small variational quantum circuit as a spectral token encoder inside a vision transformer backbone for pixel-wise land-cover mapping. A unified patch-based pipeline with band-wise normalization, principal component analysis, and quantum token encoding is evaluated on four benchmarks: Indian Pines, Pavia University, a 7-class Houston 2013 subset, and EuroSAT_MS. With roughly 35k trainable parameters, QuantFormer attains overall accuracy above 99% on the three airborne hyperspectral datasets and about 89.8% on EuroSAT_MS, competitive with deep 3D CNNs while using substantially fewer weights. Beyond full-data experiments, we also study limited-label regimes and provide practical guidance on when quantum token encoders are a viable alternative to classical projections, without claiming quantum advantage over the strongest classical baselines.
Incremental learning approach for semantic segmentation of skin histology images
(Springer Nature, 2026-03-06) Fatima, Sana; Salam, Anum Abdul; Akram, Muhammad Usman; Hameed, Ibrahim A.; Ahmed, Saad Bin
This study presents an incremental learning framework to enhance the generalization and robustness of transformer-based deep learning models for segmenting skin cancer and related tissue structures. While deep learning models often perform well on data distributions similar to their training sets, their accuracy typically degrades when exposed to novel scenarios–limiting their clinical utility in skin cancer diagnosis. To address this, we propose a biologically inspired incremental learning strategy tailored for skin cancer classification and segmentation, allowing the model to incorporate new data progressively while reducing catastrophic forgetting. Our approach integrates multiple loss functions to preserve existing knowledge while adapting to additional magnification levels. Experimental results on the indistribution test set demonstrate consistent performance improvements: achieving 89.05% accuracy with 10× magnification, 92.68% with 10× and 5× combined, and 95.53% when incorporating 10×, 5×, and 2× magnifications. These findings highlight the potential of our method to improve the adaptability and reliability of deep learning systems for empirical generalization in skin cancer classification tasks.
Laboratory- and field- scale bioassays for predicting responses of wild rice (Zizania palustris L.) to exposures of site-specific sediments and waters
(2020) Tedrow, O’Niell Roy; Lee, Peter; Leung, Kam; Kanavillil, Nandakumar; Bloom, Paul
Four bioassays were used to expose wild rice (WR) to site-specific waters and sediments within the boundaries of site-specific conditions. Mesocosm- and microcosm- scale bioassays were developed to measure responses of WR to sediment exposures. In mesocosms, WR height (HT), dry weight biomass (DWB), and seed production (SP) were statistically lower for Cleaver and Unnamed Lake sediment-grown plants compared to Rat River Bay sediment-grown plants. An accelerated-growth microcosm-scale bioassay accurately represented the mesocosm-scale bioassay, while decreasing overall time, sediment, water, and space. WR developed to near reproductive maturity. Significant differences were not identified between mesocosm:microcosm ratios for WR DWB or SP, which appeared to be primarily influenced by sediment ammonia-nitrogen concentrations. Rafts were deployed in two select aquatic systems: three in the Seine River (non-industry-influenced); and two rafts in each of three legacy-mine influenced pits to determine the life stage (aerial, floating leaf, submerged) more sensitive to water depth and water depth increases. Based on data obtained during this study, floating leaf plants were determined more sensitive to depth increases; aerial stage was least sensitive to depth increases. WR developed to aerial stage in 20 and 40 cm water depth treatments in all legacy-mine influenced pits indicating no adverse responses to pit waters. Two flow-through WR paddies were constructed adjacent to separate legacy-mine influenced pits with elevated sulphate concentrations. Seeded WR in each paddy developed according to typical phenology during each of multiple successive growing seasons. In the Pit A paddy, no statistical decreases in HT or DWB were observed between growing seasons; average SP statistically increased in 2019; and seed DWB remained statistically similar between growing seasons. Over two successive growing seasons, WR stem density remained statistically similar in the Pit C paddy. Despite conditions potentially conducive to iron sulphide root coating formation, this was not identified via SEM-EDX characterization. No adverse WR responses observed throughout this study were determined resultant of Pit A or Pit C water exposures. Representativeness of natural WR areas is paramount to bioassay data defensibility. Bioassays described herein were designed to represent field conditions to the extent possible given the scale of the bioassay.