Master Theses & Projects (FSCI)

Permanent URI for this collectionhttps://hdl.handle.net/10155/387

Browse

Now showing 1 - 20 of 344

Multi-omic analysis of contaminated sediments from Contrecoeur, QC on Planorbid snails
(2024-07-01) St-Laurent-Guérin, Jeanne; Gilroy, Ève; Simmons, Denina
Contrecoeur’s history of industrial development resulted in sediment contamination (by butyltins, heavy metals, and petroleum hydrocarbons), which can be harmful to aquatic and benthic life. Additionally, future development threatens to further exacerbate the contamination in the area. This study examines the effect of Contrecoeur’s contaminated sediment and increased water temperature on Planorbid snails’ survival, growth, reproductive output, metabolomic profile, and proteomic profile. A two-week in-situ exposure was performed on Helisoma trivolvis, and a 28-d laboratory exposure of Planorbella pilsbryi to the same sediments was completed. Snail tissue samples were collected for proteomic and metabolomic analysis. Sediment contamination caused mortality and affected the insulin signaling and the transsulfuration pathways. Thermal stress increased the effect of ammonia through urea metabolites and affected the abundances of guanidinoacetate, creatine, and creatinine. The other effects (on DNA translation and transcription, on energy metabolism, and on key biomarker proteins) cannot be attributed specifically to one stressor.
Bias and fairness in transfer learning
(2024-07-01) Salmani, Parisa; Lewis, Peter R.
Transfer learning involves using knowledge from one task to improve performance and reduce training time on a related task. However, recent studies highlight a critical issue: the fairness of models trained with transfer learning. One study showed that transfer learning can transfer intentionally planted biases from the source task to the target task. This thesis explores a different but equally critical problem: whether transfer learning can introduce new biases or lead to greater biases in the target task. Our investigation reveals that transfer learning can introduce varying degrees of bias in the target task that were not present in the source task. We examined two applications that commonly use transfer learning. Our findings indicate that, in both cases, transfer learning increases bias concerning sex, age, and race compared to non-transfer learning methods trained from scratch, which are nearly as accurate. These results emphasize the need for understanding the limitations and risks of transfer learning, especially in high risk domains like healthcare and security, and call for further research into conditions under which transfer learning introduces and amplifies bias.
A targeted reverse mapping machine learning approach for non-dominated solutions in multi-objective optimization
(2024-08-01) Kermani Poor, Masoud; Ibrahimi, Mehran; Rahnamayan, Shahryar
Multi-objective optimization problems aim to identify solutions that maximize or minimize conflicting objectives. Population-based multi-objective algorithms, inspired by biological populations, are effective but often provide limited solutions within the decisionmakers’ region of interest (ROI) on the Pareto front. Recent advancements in machine learning have shown promise in generating solutions, yet they suffer from a lack of control and require knowledge of objective function attributes. This study proposes a framework using Gaussian process regression and artificial neural networks to generate innovative solutions in the ROI. By employing diverse sampling techniques and integrating long term memory, the framework can produce more than twice as many solutions in the ROI, as demonstrated in experiments with real-world problems and various benchmark functions.
Comparison of performance for visual feedback and cursors in mid-air grasping of 3D objects
(2024-08-01) Culver, Claire Madilyn; Kapralos, Bill
While investigating the performance benefits of visual feedback for grasping in midair interaction based virtual reality simulations, some studies noted that user performance in tasks has not improved when using visual feedback methods, while it has improved in other, similar studies. This thesis presents an experiment that was conducted to investigate the effectiveness of these techniques. The experiment itself involves a simple task that uses tools through a grasp and release mechanism, in a virtual environment that compares visual feedback methods used in other studies against techniques meant to inform the user of the bounds of the simulation in advance, known as feedforward techniques. Data collected finds that performance is not significantly impacted by either feedback or feedforward techniques, though feedback is preferred against a lack of feedback. Recommendations are provided to help minimize the use of visual effects on objects that may draw user focus from their intended target.
Using non-targeted mass spectrometry to identify proteomic, metabolomic and lipidomic profiles in saliva to compare the impact of prolonged vs. interrupted sitting
(2024-08-01) Chauhan, Mohammed Faiz; Simmons, Denina
Traditional sedentary physiology research focuses on targeted approaches. To explore from a new perspective, a non-targeted approach with the use of Liquid Chromatography-Mass Spectrometry (LC-MS) to assess the responsiveness of proteins and metabolites in saliva to sedentarism was employed. The study involved 24 participants that engaged in both a prolonged sitting session and an interrupted sitting session, in a randomized order. Saliva samples were collected before and after each session, including a separate baseline. Samples were purified and concentrated then analyzed on the LC-MS. A total of 2493 proteins, 17 lipids, and 11 biogenic amines were detected. 2112 proteins were significantly differentially abundant among the different session sample groups. In conclusion, prolonged sedentary behaviour of 4 hours led to a range of molecular responses that were observed through changes in the relative abundances of proteins and lipids. The results were complex and suggestive of several indicators of health deterioration.
Detection of trace concentrations of small and big molecules through vibrational spectroscopy
(2024-08-01) Balasubramanian, Janani; Agarwal, Nisha Rani
In this thesis, we aim to detect analytes of varying sizes using surface-enhanced Raman spectroscopy (SERS) without the need for customizing the morphology of nanostructured substrate. Signal enhancement in SERS occurs near plasmonic nanoparticles within "hotspots" (<10 nm), which limits detection to molecules that can access these regions. We investigated two differently fabricated SERS-active substrates: pulsed laser deposited gold nanoparticles and electrochemically deposited silver nanoparticles, optimized for 633 nm and 532 nm Raman lasers, respectively. Characterization was performed using UV-Visible spectroscopy, scanning electron microscopy, X-ray diffraction, and X-ray photoelectron spectroscopy. We analyzed zeatin, a small plant hormone (219 Da), and hemoglobin (Hb), a large protein (64,500 Da). Zeatin detection ranged from 1 mM to 1 nM, even in complex bacterial media, while Hb detection involved ligand-functionalized substrates targeting the heme group for specific oriented immobilization, detecting down to 10 nM. This study demonstrates the potential of SERS for sensitive and selective detection of diverse analytes, paving the way for advanced biosensing applications.
Staffing queueing systems with cyclical demand and unreliable servers
(2024-08-01) Amini, Abraham; Rastpour, Amir
Unplanned staff absences, referring to scheduled servers being unavailable, pose significant challenges in sectors like healthcare, airlines, and correctional facilities, leading to under-staffing and management reliability issues. This thesis investigates the effectiveness of the Stationary Independent Period by Period (SIPP) method for staffing a multi-server delay queueing system with time-varying demand and server absence. Using the M(t)/M/c(t) queueing model, we systematically examine the SIPP method across various scenarios, including realistic ones, considering cyclical customer arrival rates and multiple servers with uncertain availability. We identify the parameter settings under which the SIPP method is most compromised. We propose two modifications to the SIPP method to account for absence in staffing decisions. We systematically compare these two proposed methods across different scenarios and identify the parameter settings under which each proposed method performs better. We also propose a heuristic to schedule additional servers given a certain budget.
Building reliable AI assistive technologies: a comprehensive process for inclusive and transparent co-design
(2024-08-01) Ahmadi, Zahra; Lewis, Peter
Artificial Intelligence (AI) is used to improve assistive technologies, but these systems can fail in various ways. Despite being designed for individuals, many technologies lack user engagement from design to evaluation. This raises the question for users: “Can I trust this technology to perform its intended task?” We conducted a Systematic Literature Review (SLR) of AI-based assistive technologies for persons with visual impairments, focusing on how studies report potential risks and failures. Our findings reveal that many systems lack evaluation by the sight-loss community, and many studies do not adequately report failure cases and associated risks. This oversight can lead to serious safety concerns. To address these issues, we proposed TACTIC: a process for Transparent, Accessible, Co-design, Through Inclusive, Iterative Cycles, emphasizing iterative end-user engagement. TACTIC includes four co-design loops: problem identification, methodology design, solution evaluation, and knowledge sharing. This process aims to improve system design, community engagement, and the standardized reporting of risks, thereby enhancing the safety and effectiveness of AI-based assistive technologies.
A study of meta-learning methods on the problem of video matting
(2024-04-01) Tabaraki, Negin; Qureshi, Faisal; Pu, Ken
Applying image matting techniques directly to video matting presents challenges, primarily due to the complex temporal dynamics inherent in video data. In this work, we studied two Meta Learning approaches—Boosting with Adapters (BwA) and Boosting using Ensemble (BuE)—to tackle the task of video matting using pre-trained image matting models. BwA refines (image matting) alpha mattes by fine tuning pre-trained segmentation models, which we refer to as adapters, using video frames. BuE, additionally, combines multiple fine-tuned adapters using a convolutional neural network. We introduced a meta-learning architecture that incorporates both adapters and ensemble boosting through an iterative process of expert selection and fine tuning. Based on our evaluation on benchmarks based on a standard video matting dataset (VideoMatte240K), we confirm that the proposed scheme improves the performance of image matting models on the task of video matting. In addition, the proposed approach also improves the performance of VMFormer (c. 2022), a recent video matting method.
Artificial intelligence agent for assessing game difficulty
(2024-04-01) Sansalone, Stevie C. F.; Mirza-Babaei, Pejman
Balancing game difficulty is a key element of developing games which engage players without causing frustration. Difficulty balancing is an iterative process and game developers frequently rely on user tests to guide them, but the cost of user testing can be prohibitive, particularly for independent developers. In response to this, I seek to determine whether AI can be used to test the difficulty of developing games before setting up a full playtest. For this purpose, I developed ARTemiS, a prototype tool for AI-based playtesting focused on game difficulty arising from precision input tasks such as combat. I then performed a user study with 10 participants who used the tool to assess and modify three demo levels. This research seeks to assess the applicability and utility of AI-based playtesting for difficulty balancing and lay a foundation for the development of tools for assessing other aspects of difficulty with AI.
Quadruped robots and the uncanny valley: a study on canine representation
(2024-04-01) Padilla Velasco, Carolina; Hung, Patrick
This research analyses how dog-like features in robotic quadrupeds influence their social perception. The initial part of the methodology describes the design and development of three robotic dog prototypes that include key features like a head and a tail. Further, Contrastive Language-Image Pre-Training (CLIP), a neural network that has demonstrated signatures of the uncanny valley effect before, was used to explore how the perception of quadrupeds evolves as their level of canine likeness intensifies. For this purpose, seven models were tested, ranging from a fully robotic quadruped to a living dog, and 252 images were assessed for each. Our findings indicate that the uncanny valley effect also develops in quadruped robots. This novel contribution serves as a reference to select an appropriate level of realism for four-legged robots. This is particularly valuable for robotic applications that look to incorporate human-dog dynamics.
Robotic tails in education for enhanced Human-Robot Interaction
(2024-04-01) Martinez Cano, Marco Antonio; Hung, Patrick
This thesis presents an articulated tail mechanism to enhance quadruped robots, enabling them to mimic expressive behaviors like dogs. Additionally, this work investigates non-verbal communication in quadruped robots for Human-Robot Interaction (HRI) while exploring the potential of utilizing a robotic tail prototype as an education tool in academia, introducing Science, Technology, Engineering and Mathematics (STEM) students to the development of social robotics. This thesis describes the development of the Do It Yourself (DIY) Robotic Tail Workshop for the “Special Topics in Information Technology-Service Robots + Topics in Technology Management-Service Robots” class at Ontario Tech University, listing the objectives, methods, and considerations presented through the three times the workshop has been executed. Furthermore, an experiment and comparative analysis were conducted using the first and current tail prototypes by performing an accuracy and repeatability parameter testing procedure based on the Robot-Test and Ford methods. The experimental results indicate higher levels of accuracy and a broader range of motion for the latest tail prototype compared to similar commercial products available in the market, such as the Qoobo robot, a companion cushion robot with a moving tail that responds to physical touch. This thesis contributes to establishing robotic tails as an innovative approach to HRI by setting the basis to bring emotional expression to robots.
Run, Llama, Run: an educational coding game for assessing tangible and hybrid interfaces
(2024-04-01) Koornneef, Stacey A.; Bradbury , Jeremy; Miljanovic, Michael
As the impact of computer science becomes more pervasive across the world, it is increasingly being integrated into curriculum at earlier and earlier grades, including from kindergarten through grade 5 (which is the grades of early elementary school). One of the important concepts in computer science at earlier ages is computational thinking, or the process of how to think algorithmically. It is important when developing new ways to teach computer science that they are collaborative, engaging, and cost-accessible. This thesis proposes a new cost-accessible educational game for K-5 students that teaches computational thinking through an educational coding game called Run, Llama, Run. This game has both fully tangible and hybrid digital/tangible versions, which allow us to assess the impact of the interfaces on engagement and collaboration. Its purpose is to determine the benefits and detriments of a fully tangible educational game for this age group, focusing on collaboration and engagement.
Characterization of apoptotic cell death in bovine red blood cells
(2024-04-01) Kennedy, Bailey; Qadri, Syed
Red blood cells (RBCs) in mammalian species display phenotypic variations in cellular functions and metabolism ascribed to their compositional and morphological differences. After accruing stress-induced damages, organelle-free human RBCs display a specialized apoptotic cell death process characterized by breakdown of normal phospholipid cell membrane architecture. This process facilitates swift phagocytic recognition and catabolism of apoptotic RBCs, which could significantly curtail their lifespan in circulation. Due to multiple phospholipid anomalies in bovine RBCs, their cell death machinery is not completely understood. Herein, we observed that bovine RBCs display differential cell death patterns in response to various pathophysiologic cell stressors in vitro as compared to human RBCs, which were only partly explained by increased Ca2+ influx and oxidative stress. In conclusion, premature cell death of circulating bovine RBCs could potentially contribute to the pathogenesis of anemia in cattle of varying etiology. The present observations may have relevance to livestock health and productivity.
An investigation of the environmental factors that affect water quality and the occurrence of harmful cyanobacteria in stormwater management ponds
(2024-04-01) Horton, Kaitlyn; Kirkwood, Andrea
Stormwater management ponds (SWMP) have an important role in flood mitigation and basic water quality treatment via sedimentation. As aquatic ecosystems, less is known about their role as habitats for aquatic organisms and their potential to transform pollutants. This study focused on twelve SWMP in Oshawa, Ontario, Canada to assess the impacts of flow conditions and pond characteristics on SWMP water quality treatment and algal growth, including toxic cyanobacteria. Net release of total phosphorus (TP) and net retention of total nitrogen (TN) by SWMP were observed. Flow conditions had little affect on the overall functioning of the study SWMP. Large pond designs and recent dredging were observed to positively influence the reduction of total suspended solids (TSS), TN, and TP net release from SWMP. Algae and cyanobacteria were observed to be generally N-limited. The presence of the cyanobacterial toxin gene mcyE was positively associated with chloride and heterocyst-forming cyanobacteria.
A few-shot learning method for single-object visual anomaly detection
(2024-04-01) Ejaz, Neha; Qureshi, Faisal
We propose a few-shot learning method for visually inspecting single objects in an industrial setting. The proposed method is able to identify whether or not an object is defective by comparing its visual appearance with a small set of images of the “working” object, i.e., the object that passes the visual inspection. The method does not require images of defective objects. Furthermore, the method does not need to be “trained” when used to inspect new, previously unseen, objects. This suggests that the method can be easily deployed in industrial settings. We have evaluated the method on three visual anomaly detection benchmarks—1) MVTec, 2) MPDD, and 3) VisA. On the first two datasets the proposed method achieves performance that is comparable to state-ofthe- art methods that require access to object-specific training data. Model performance on VisA is poor; however, it is to be noted that the model was never trained on VisA dataset. We also show that the proposed model boasts fast inference times, which is a plus for industry applications. This project is funded in part by Axiom Plastics Inc., and we have evaluated the proposed method on a proprietary dataset provided by Axiom. The results confirm that the proposed method is well-suited for single-object visual anomaly detection in industry settings.
Characterizing midair handwriting in virtual reality
(2024-04-01) Chan, Matthew; Collins, Christopher; Qureshi, Faisal
Midair handwriting poses challenges due to the lack of a physical plane to press against while writing, making it difficult to determine when ink should be placed. In this thesis, we gathered midair handwriting data from 24 participants in an environment that allowed them to write freely. We compared writing with a pen-like object and writing using a finger across two writing methods (writing freely versus on a virtual whiteboard). Using our data, we trained a neural network to detect when ink should be placed during midair handwriting, achieving an overall 85% accuracy. We developed a data-viewing application to recreate sentences for visual analysis. Participant feedback favoured the pen-like object as a writing utensil, with equal preference for both writing methods. Our contributions include a midair handwriting Virtual Reality (VR) application for data collection, a dataset containing 480 sentences of frame-by-frame midair handwriting data, and 20 unique prompts used in participant trials.
Generative methods for image synthesis with applications to medical imaging
(2024-04-01) Abdollahi, Melika; Ebrahimi, Mehran; Davoudi, Heidar(Koroush)
There has been significant advancement in the field of medical image synthesis for diagnostic and analytical improvements. Generative methods for image synthesis have been an active area of research in recent years. In this thesis, we explore the use of a hybrid model of generative adversarial networks (GANs) and transformer networks for image synthesis. We propose a method that combines GANs with transformer networks to address the translation and super-resolution of medical images. We also present a model for inpainting of images. Finally, we introduce a re-parameterization model that translates one image modality to another modality with paired parameters. The proposed methods are evaluated qualitatively and quantitatively. Our results show that our proposed models are feasible and applicable for super-resolution, translation, inpainting and re-parameterization.
Preliminary design of a social asymmetric virtual reality upper-limb exergame for individuals with dementia utilizing insights from caregivers
(2023-10-01) Saunders, Stephen; Quevedo, Alvaro
This thesis investigates the usability, task load, cooperative performance, and social presence effects of an asymmetric VR game for upper limb activity in the context of elderly care from the perspective of caregivers. The study presented participants with three different play modes: Cooperative within immersive VR, cooperative external to VR, and single-player within VR. The results indicate that the three conditions had above-average usability and social presence and a task load score lower than that of average daily activities. Additionally, a Sign test between the cooperative versions revealed a statistically significant difference in mean Behavioural Engagement scores favouring the version external to VR, p = 0.031. Although future studies with larger sample sizes are needed for an effective evaluation, these results indicate the exergame shows much promise in providing a highly usable, low cognitive load, socially involved exergame for people with dementia and their caregivers.
Exploring volumetric video & VR on self-efficacy for first aid training - a pilot study
(2023-12-01) Orian, Colin; Hogue, Andrew
The health profession is currently in a global crisis due to the lack of health professionals, such as nurses and doctors. In response to this ongoing crisis, extended reality is being investigated as a potential modality for teaching the next generation of health professionals. In addition to extended reality being used for teaching, dynamic recordings of sequential 3-dimensional models, also known as volumetric videos, have been investigated for their use in education. However, there is a limited amount of research on how volumetric videos compare to conventional 2D videos. Therefore, this thesis compares how volumetric videos and 2D videos influence a person’s self confidence by having participants learn how to perform head bandaging in virtual reality through watching either video type. A significant difference in self confidence was found after viewing an instructional video on head bandaging. A significant difference in presence between the videos was also found.

Browse

Recent Submissions