Digital Commons @ Florida Atlantic
Not a member yet
400 research outputs found
Sort by
A GENERATIVE AI PIPELINE FOR RARE ROAD SCENARIO AUGMENTATION FROM EXISTING TRAINING DATASETS
No full textAutonomous Vehicle (AV) systems have the potential to transform the transportation ecosystem significantly by improving safety and accessibility. One of the core components of any AV is a real-time object detection system. It helps an AV localize its position in the surrounding environment and perform tasks such as maneuver planning and obstacle detection. To make a deep-learning-based object detection system safe and robust, it must be trained on diverse and rare scenarios that can occur in road conditions. However, it is expensive and time consuming to manually create training datasets containing rare road objects or conditions. Synthetic training data can be a practical solution to this problem.
This thesis proposes a Generative AI pipeline that takes images from an existing training dataset, identifies a suitable region in each image to create an inpainting mask, places a desired rare road-scenario object via text-guided inpainting, and generates annotations of the inpainted object within the placement mask. In this work, I implemented the proposed Generative AI pipeline to inpaint vehicles on fire into road images from the NuScenes dataset and generated bounding-box annotations of the fire regions in YOLO format.
To evaluate the e!ectiveness of the synthetically generated images, I conducted an experiment by training two YOLO models from Ultralytics: one using a real world training dataset and another using synthetically generated inpainted images. Both models were tested on real world images of vehicles on fire, sourced from the internet and manually annotated by me. The YOLO model trained with synthetic images demonstrated comparable performance to the benchmark model trained with real world data, showing that synthetic images can substantially reduce the challenges of curating training datasets for rare road scenarios. Furthermore, I discussed the challenges faced while implementing the synthetic data generation pipeline, such as the complexity of identifying and annotating fire in images
WAR ABSTRACTED
No full textWar Abstracted is an exhibition in response to the devastation of war. The body of work encapsulates three years of artistic exploration at Florida Atlantic University. The exhibit is a reaction to contemporary conflicts, both global and domestic. It is neither a glorification nor a condemnation of war. The exhibition reflects my experience as a Marine Corps Scout during the Gulf War, the challenges of reintegrating into society with post-traumatic stress disorder (PTSD), and uncertainty. It also engages broader themes of global warming and gun violence. Through a diverse range of materials, methods, and presentations, the work navigates personal and political landscapes, occasionally voicing a decisive perspective on divisive issues
COMPARATIVE ANALYSIS OF 2D AND 3D GAIT METRICS FROM VIDEO USING DEEP LEARNING TECHNIQUES
No full textThis thesis presents a comparative evaluation of two deep learning-based methods for extracting 2D and 3D gait metrics from monocular video: a 2D-to-3D lifting approach combining AlphaPose for keypoint detection with MotionBERT for temporal 3D reconstruction, and a direct 3D estimation method using MeTRAbs. Videos of two healthy adults walking across a ProtoKinetics Zeno™ Walkway were captured laterally at 4K/60 fps using an iPhone 16. Ground-truth spatiotemporal gait parameters were obtained from the pressure-sensitive mat, enabling quantitative validation via Mean Absolute Error (MAE) and Pearson correlation. MeTRAbs consistently outperformed the lifting pipeline, achieving MAE below 2% for temporal metrics (e.g., 1.58% for step time, 1.09% for gait cycle time) and strong correlations (r \u3e 0.90, p \u3c 0.001). Spatial metrics showed MAE of 15.28% (step length) and 15.84% (stride length), with superior robustness to occlusions. AlphaPose+MotionBERT exhibited higher spatial errors (20.40% and 18.47%, respectively) and weaker correlations, though it remained viable for velocity (9.34% MAE) and low-resource settings. The results highlight direct 3D estimation as the preferred method for clinical precision and biomechanical fidelity, while 2D-to-3D lifting offers a lightweight alternative for scalable, non-intrusive monitoring. Future work should expand to pathological gaits, multi-view fusion, and edge-optimized models to broaden real-world applicability
EFFICIENT AND ROBUST INVERSE REINFORCEMENT LEARNING FOR COMPLEX AND DYNAMIC ENVIRONMENTS
No full textWith the rapid progress of reinforcement learning (RL) and its applications in robotics, autonomous driving, and energy systems, learning reliable reward functions has become a central challenge. Traditional RL relies heavily on handcrafted reward functions, which are often infeasible to design in complex real-world environments. Inverse reinforcement learning (IRL) provides an alternative by recovering reward functions from expert demonstrations. However, existing IRL methods suffer from inefficiency, poor robustness under perturbations, and limited generalization to unseen scenarios.
This dissertation focuses on advancing IRL from three complementary perspectives: computational efficiency, robustness, and generalization. Specifically, the following problems are investigated: 1. Computational Efficiency: We develop a novel, efficient IRL (e-IRL) framework that reformulates reward recovery through feature expectation matching, eliminating the need for state visitation estimation. This reduces memory and computation costs while scaling effectively to high-dimensional environments. 2. Robustness: To address perturbed environments, we introduce a multi-virtual-agent IRL (MVIRL) framework. By training multiple agents across parallel perturbed environments with weight-sharing and data aggregation, MVIRL learns robust reward functions that remain stable under noise, gravity variations, and adversarial conditions. 3. Generalization and Personalization: We propose a contrastive IRL (CIRL) framework that integrates self-supervised contrastive representation learning with maximum entropy IRL. CIRL leverages momentum encoders and reward-regularized contrastive loss to improve sample efficiency and adapt to personalized driving styles, ensuring both robustness and adaptability
FETAL MICROCHIMERISM OF THE EYE LENS
No full textFetal microchimerism (FMc) is the presence of a small fetal cell population in the mother, sometimes exhibiting stem cell–like properties. FMc has been observed within many placental mammal organs, except the eye. We used Yspecific PCR on genomic mouse DNA to test the hypothesis that fetal microchimerism can also be found in the lens. Of the parous females tested, we detected an Ssty2-specific amplicon in 83% (5/6) of samples and an Sry-specific amplicon in 2/4 samples. 100% (5/5) of non-age-matched parous females were positive for both primers. These results provide evidence that cells of fetal or placental origin can migrate to the maternal mouse lens, and plans to investigate these cells\u27 potential to differentiate into lens or immune cells are underway
LOBIZONA AND CEMETERY BOYS: BROADENING REPRESENTATION OF THE LATINX/CHICANX COMMUNITY
No full textThis thesis focuses on two young adult speculative fiction novels, Lobizona and Cemetery Boys, that showcase the need for accurate representation of the Latinx/Chicanx and LGBTQ+ communities. Themes of sexuality, race, identity, and the gender binary within Latinx/Chicanx cultures will be explored. The narratives of the protagonists in both novels disrupt the gender binary in the United States and the Latinx/Chicanx community. There is a lack of literature that focuses on Latinx/Chicanx young adult speculative fiction, particularly in the United States, highlighting the need for more scholarly research in this area. It is essential for literary scholars and readers to give recognition to the existence of these novels and their cultural significance for marginalized communities
CONSERVATION ECOLOGY OF THE WHITESPOTTED EAGLE RAY (AETOBATUS NARINARI) IN FLORIDA’S COASTAL WATERS
No full textElasmobranchs (sharks, skates, and rays) are among the most threatened vertebrate groups due to biological and anthropogenic pressures. Their low fecundity, late maturity, and slow growth make them susceptible to overfishing, habitat degradation, and habitat loss. A trade-off for the low fecundity is the investment in juvenile survival through the use of nurseries, which can play a crucial role in population recovery. However, these nurseries, often located in shallow coastal areas, are increasingly degraded by coastal development and are frequently utilized for human activities, leading to conflicts between humans and wildlife.
This study combined acoustic telemetry, phycotoxin analysis, and a citizen science approach to provide information for the conservation of the globally endangered whitespotted eagle ray (Aetobatus narinari) in Florida and beyond. First, this study aimed to identify nursery areas for A. narinari in Florida to aid with population recovery (Chapter 1). Moreover, this study gathered information on the uptake of phycotoxins (Chapter 2) and assessed negative interactions with rays in the shellfish aquaculture industry (Chapter 3). Acoustic telemetry supported the role of Sarasota Bay and the Indian River Lagoon (IRL) as nurseries with young-of-the-year rays remaining in these areas for extended periods. Phycotoxin analysis confirmed the uptake of phycotoxins across life stages and tissues, highlighting dietary exposure as a primary pathway. A survey analysis confirmed the perception of sharks and rays as a safety threat to Florida’s shellfish farmers. Stingrays and cownose rays were reported as the primary species interacting with the gear, while interactions with A. narinari remained unclear. Captive trials demonstrated a 29% reduction in feeding success with the Sharkbanz Zeppelin and eliminated bag damage, showing a promising avenue to explore non-lethal deterrents as a mitigation strategy.
This study provided evidence for the use of Sarasota Bay and the IRL as eagle ray nurseries, highlighting harmful algal blooms (HAB) as a key conservation challenge. These findings underscore the need to accelerate research on HAB mitigation strategies to preserve ecologically important estuaries sustaining eagle ray populations and human activities. Equally important is the exploration of nonlethal deterrents in shellfish aquaculture to foster coexistence between rays and humans
CHALLENGING TRUTH, EXPLORING FLUIDITY, AND RECLAIMING BODILY AUTONOMY THROUGH REVISITED PERSONAL NARRATIVES
No full textThis dissertation explores and analyzes revisited personal narratives, which I define as layering perspectives or theorizing upon previously presented narratives. Specifically, I argue that revisited personal narratives are uniquely positioned to not only explore the intricacies of fluidity and truth within narrative but also deepen the insight into the perspectives of marginalized identities, thereby aiding in the deconstruction of cisheteronormativity and the reclaiming of autonomy—both bodily and narrative. Bodily autonomy refers to the freedom and ability to make choices about one’s own body and identity, including the right to not have one’s body violated, the right to access healthcare, the right to live freely, and the right to basic human rights, and narrative autonomy refers to the right to hold ownership over personal stories, experiences, and identities, including choosing when, how, and with whom to share those experiences. I explore numerous revisitation types throughout this dissertation. For example, in Chanel Miller’s memoir, Know My Name, she juxtaposes three distinct narrative perspectives. In doing so, she uncovers the effect of narrative presentation in interpretation while simultaneously challenging rape myths and reclaiming her identity. Alternatively, Janet Mock navigates fluidity and truth within identity by discussing her narrative decisions within her two memoirs, Redefining Realness: My Path to Womanhood, Identity, Love & So Much More and Surpassing Certainty: What My Twenties Taught Me. Her narrative decisions correlate to her challenging of societal assumptions regarding trans identities, particularly trans women of color. To further engage with these narrative tactics, I revisit my master’s thesis through an examination of my narrative choices and presentations. The dissertation concludes with a demonstration of how revisited personal narratives can be actively engaged to foster greater understanding of oppressed identities. In doing so, I demonstrate how revisited personal narratives are uniquely positioned to not only serve as an act of resistance against cisheteronormativity but also reclaim bodily autonomy
FOOD SOVEREIGNTY AND ADVOCACY COALITIONS IN THE UNITED STATES
No full textThis dissertation examines how food sovereignty advocacy coalitions in the United States shape beliefs, conduct advocacy work, influence discourse, and perceive their policy influence. The original definitions of food sovereignty emphasize producers and control of the food system, but in the U.S., it has been adapted to incorporate equality, Indigenous sovereignty, and local food movements. However, little research has considered the role advocacy coalitions play in the food sovereignty policy process.
Using the Advocacy Coalition Framework, this study examines how food sovereignty advocacy groups define the food sovereignty problem, coordinate actions, and attempt to influence the policy process. A qualitative document analysis of 64 documents from 2003 to 2024 and five interviews with food sovereignty advocates provides the data for coding belief systems, advocacy strategies, food sovereignty discourse, and perceived effectiveness.
The findings of this dissertation show that food sovereignty’s deep core beliefs are centered on the human right to food, while policy core beliefs and secondary beliefs are more suited for the specific U.S. context. Various advocacy strategies are used, but the most common are education, increasing organization capacity, and coalition-building, while the least common are direct actions like protests. While there are numerous variations of food sovereignty, the U.S. discourse moves away from the producer focus and instead emphasizes systemic inequities, community, and culturally appropriate foods. Overall, coalitions perceive their policy influence effectiveness to be done through small, incremental changes and are focused on growing awareness of the problem despite political threats and limited resources.
This dissertation contributes to the ACF by changing the way it has been applied in the past. This study focuses on one coalition to demonstrate the significance of secondary beliefs, discourse variations, policy-oriented learning, and temporal venue shopping. Additionally, the public administration literature is expanded by explaining how advocacy coalitions are critical components of grassroots participatory governance. The findings of this study are useful for policymakers and advocates attempting to influence food policies throughout the U.S
