{"id":{"repo_id":"alicante","oai_identifier":"oai:rua.ua.es:10045/167993"},"canonical_url":"https://search.dev.ndltd.org/etd/alicante/oai:rua.ua.es:10045/167993","repository":{"repo_id":"alicante","name":"University of Alicante","base_url":"https://rua.ua.es/server/oai/request"},"display":{"title":"Synthetic Generation and Reconstruction of Human-Centred Virtual Environments","abstract":"This thesis explores the generation of realistic virtual environments through a hybrid approach that combines traditional graphics engines with state-of-the-art generative deep learning models. The proposed methodology enables the creation of complex, adaptive, and human-centred environments for applications in cognitive assessment, synthetic video generation, and 3D object reconstruction. The first contribution is the development of an Augmented Reality (AR) application designed to assess cognitive and motor abilities through interactive tasks. Implemented on the HoloLens 2, the system captures behavioural data while offering a non-intrusive and engaging interface, particularly useful for early detection of cognitive decline in elderly populations. The second contribution introduces a synthetic video generation framework that leverages Unreal Engine 5, Gaussian Splatting, and text-guided generative motion models to simulate human actions in photorealistic virtual scenes. This approach is particularly valuable for generating data on rare or ethically sensitive events, such as elderly falls, and has demonstrated effectiveness in improving the generalisation of action recognition models through data augmentation. The final contribution presents Obj3Dify, a pipeline for reconstructing high-quality 3D models from single RGB images of handheld objects. The method integrates segmentation, diffusion-based inpainting, and neural mesh generation to overcome occlusion challenges, producing realistic object representations for integration into virtual environments. Collectively, these contributions advance the field of virtual environment generation by enabling scalable, flexible, and privacy-preserving synthetic data solutions. The research supports applications in healthcare, simulation, and content creation, contributing to the broader computer-vision, human-centric virtual systems.","abstract_html":"This thesis explores the generation of realistic virtual environments through a hybrid approach that combines traditional graphics engines with state-of-the-art generative deep learning models. The proposed methodology enables the creation of complex, adaptive, and human-centred environments for applications in cognitive assessment, synthetic video generation, and 3D object reconstruction. The first contribution is the development of an Augmented Reality (AR) application designed to assess cognitive and motor abilities through interactive tasks. Implemented on the HoloLens 2, the system captures behavioural data while offering a non-intrusive and engaging interface, particularly useful for early detection of cognitive decline in elderly populations. The second contribution introduces a synthetic video generation framework that leverages Unreal Engine 5, Gaussian Splatting, and text-guided generative motion models to simulate human actions in photorealistic virtual scenes. This approach is particularly valuable for generating data on rare or ethically sensitive events, such as elderly falls, and has demonstrated effectiveness in improving the generalisation of action recognition models through data augmentation. The final contribution presents Obj3Dify, a pipeline for reconstructing high-quality 3D models from single RGB images of handheld objects. The method integrates segmentation, diffusion-based inpainting, and neural mesh generation to overcome occlusion challenges, producing realistic object representations for integration into virtual environments. Collectively, these contributions advance the field of virtual environment generation by enabling scalable, flexible, and privacy-preserving synthetic data solutions. The research supports applications in healthcare, simulation, and content creation, contributing to the broader computer-vision, human-centric virtual systems.","abstract_has_math":false,"creators":["Mulero Pérez, David"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025","date_published":"2025","updated_at":"2026-07-24T00:52:27Z","subjects":["Synthetic Data Generation","Artificial Intelligence","Extended Reality"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["hdl:10045/167993"],"render_values":[{"text":"hdl:10045/167993","href":null,"code":true}]}]},"links":{"outbound_url":null,"outbound_label":null,"outbound_source":null},"metadata_groups":[{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2025"]},{"key":"dc:type","label":"Dc Type","values":["info:eu-repo/semantics/doctoralThesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Synthetic Data Generation","Artificial Intelligence","Extended Reality"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["hdl:10045/167993"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.other","label":"Dc Description Other","values":["This thesis explores the generation of realistic virtual environments through a hybrid approach that combines traditional graphics engines with state-of-the-art generative deep learning models. The proposed methodology enables the creation of complex, adaptive, and human-centred environments for applications in cognitive assessment, synthetic video generation, and 3D object reconstruction. The first contribution is the development of an Augmented Reality (AR) application designed to assess cognitive and motor abilities through interactive tasks. Implemented on the HoloLens 2, the system captures behavioural data while offering a non-intrusive and engaging interface, particularly useful for early detection of cognitive decline in elderly populations. The second contribution introduces a synthetic video generation framework that leverages Unreal Engine 5, Gaussian Splatting, and text-guided generative motion models to simulate human actions in photorealistic virtual scenes. This approach is particularly valuable for generating data on rare or ethically sensitive events, such as elderly falls, and has demonstrated effectiveness in improving the generalisation of action recognition models through data augmentation. The final contribution presents Obj3Dify, a pipeline for reconstructing high-quality 3D models from single RGB images of handheld objects. The method integrates segmentation, diffusion-based inpainting, and neural mesh generation to overcome occlusion challenges, producing realistic object representations for integration into virtual environments. Collectively, these contributions advance the field of virtual environment generation by enabling scalable, flexible, and privacy-preserving synthetic data solutions. The research supports applications in healthcare, simulation, and content creation, contributing to the broader computer-vision, human-centric virtual systems."]},{"key":"dc:title","label":"Title","values":["Synthetic Generation and Reconstruction of Human-Centred Virtual Environments"]}]}],"canonical_facts":{"dc:date.issued":["2025"],"dc:description.other":["This thesis explores the generation of realistic virtual environments through a hybrid approach that combines traditional graphics engines with state-of-the-art generative deep learning models. The proposed methodology enables the creation of complex, adaptive, and human-centred environments for applications in cognitive assessment, synthetic video generation, and 3D object reconstruction. The first contribution is the development of an Augmented Reality (AR) application designed to assess cognitive and motor abilities through interactive tasks. Implemented on the HoloLens 2, the system captures behavioural data while offering a non-intrusive and engaging interface, particularly useful for early detection of cognitive decline in elderly populations. The second contribution introduces a synthetic video generation framework that leverages Unreal Engine 5, Gaussian Splatting, and text-guided generative motion models to simulate human actions in photorealistic virtual scenes. This approach is particularly valuable for generating data on rare or ethically sensitive events, such as elderly falls, and has demonstrated effectiveness in improving the generalisation of action recognition models through data augmentation. The final contribution presents Obj3Dify, a pipeline for reconstructing high-quality 3D models from single RGB images of handheld objects. The method integrates segmentation, diffusion-based inpainting, and neural mesh generation to overcome occlusion challenges, producing realistic object representations for integration into virtual environments. Collectively, these contributions advance the field of virtual environment generation by enabling scalable, flexible, and privacy-preserving synthetic data solutions. The research supports applications in healthcare, simulation, and content creation, contributing to the broader computer-vision, human-centric virtual systems."],"dc:identifier":["hdl:10045/167993"],"dc:subject":["Synthetic Data Generation","Artificial Intelligence","Extended Reality"],"dc:title":["Synthetic Generation and Reconstruction of Human-Centred Virtual Environments"],"dc:type":["info:eu-repo/semantics/doctoralThesis"]},"updated_at":"2026-07-24T00:52:27Z"}