The purpose of this theoretical thesis is to research the problem: producing optimized content for handheld real-time CG applications. This thesis is based on existing literature studiesregarding the subject. It will explore the topic on how to deal with draw calls, shaderoptimizations, hard/smooth edges and how engines deal with polygons and textures and howthis correlates with current technical limitations of current game engines for handheld realtime applications such as the iPhone. On these grounds, the thesis will expand on the subject of how those topics affect the hardware; more specifically targeted the GPU and CPU onmobile devices. Today, the hardware used on mobile devices on the market is limited; it is important tounderstand that a lot of factors come in to play and making optimization a part of the design, not a final step regarding the creation of content is vital when creating CG applications formobile devices. A closer look at the available evidences suggests that when dealing with the limited resources available every performance aspect should be closely looked into. It is vital to optimize the content and correlate your game budget accordingly to the marketed device. In the end it will make for a better game. This thesis sheds a light on the development strategiesfor CG applications on handheld devices.

Our goal with this study is to examine how much optimization of 3D game models can affect the overall performance of a game. After a previous pilot study we decided on use a method where we worked with a 3D scene which was made earlier unconnected to this study. We created two versions of the scene in Unreal Development Kit, one with none of the meshes optimized and the second scene where the meshes are optimized. From these two scenes we wrote down the different stats: the draw calls, frame rate, millisecond per frame and visible static mesh elements as well as the memory usage. Comparing these stats from the two scenes, we found that there was a change in the stats. Draw calls and frame rate had dropped in the second scene, as well as the memory usage which made the game run more smoothly without losing much of its aesthetic quality. / Målet med vår studie var att se hur stor skillnad optimering av 3D-modeller i spel gör för att förbättra spelprestandan. Efter att ha utfört en pilotstudie beslutade vi oss för att använda en tidigare byggd 3D-scen för undersökningen i vår C-uppsats. Vi skapade två versioner av scenen i Unreal Development Kit, en där inga modeller var optimerade och den andra där vi optimerat modellerna. Vi skrev därefter ner statistik från de olika scenerna, nämligen draw calls, frame rate, millisecond per frame och visible static mesh elements liksom minnesanvändning. Efter att ha jämfört resultaten såg vi att det fanns en väsentlig skillnad mellan scenerna prestandamässigt. Både draw calls, frame rate och minnesanvändningen hade minskat efter optimeringen vilket ledde till att spelet kördes smidigare.

Interactions gestuelles multi-point et géométrie déformable pour l'édition 3D sur écran tactile: Malgré les progrès en capture d'objets réels et en génération procédurale, la création de contenus pour les mondes virtuels ne peut se faire sans interaction humaine. Cette thèse propose d'exploiter les nouvelles technologies tactiles (écrans "multi-touch") pour offrir une interaction 2D simple et intuitive afin de naviguer dans un environnement virtuel, et d'y manipuler, positionner et déformer des objets 3D. En premier lieu, nous étudions les possibilité et les limitations gestuelles de la main et des doigts lors d'une interaction sur écran tactile afin de découvrir quels gestes semblent les plus adaptés à l'édition des environnements et des objets 3D. En particulier, nous évaluons le nombre de degré de liberté efficaces d'une main humaine lorsque son geste est contraint à une surface plane. Nous proposons également une nouvelle méthode d'analyse gestuelle par phases permettant d'identifier en temps réel les mouvements clés de la main et des doigts. Ces résultats, combinés à plusieurs études utilisateur spécifiques, débouchent sur l'identification d'un patron pour les interactions gestuelles de base incluant non seulement navigation (placement de caméra), mais aussi placement, rotation et mise à l'échelle des objets. Ce patron est étendu dans un second temps aux déformations complexes (ajout et suppression de matière ainsi que courbure ou torsion des objets, avec contrôle de la localité). Tout ceci nous permet de proposer et d'évaluer une interface d'édition des mondes 3D permettant une interaction tactile naturelle, pour laquelle le choix du mode (navigation, positionnement ou déformation) et des tâches correspondantes est automatiquement géré par le système en fonction du geste et de son contexte (sans menu ni boutons). Enfin, nous étendons cette interface pour y intégrer des déformations plus complexe à travers le transfert de vêtements d'un personnage à un autre, qui est étendu pour permettre la déformation interactive du vêtement lorsque le personnage qui le porte est déformé par interaction tactile. / Multi-touch gesture interactions and deformable geometry for 3D edition on touch screen: Despite the advances made in the fields of existing objects capture and of procedural generation, creation of content for virtual worlds can not be perform without human interaction. This thesis suggests to exploit new touch devices ("multi-touch" screens) to obtain an easy, intuitive 2D interaction in order to navigate inside a virtual environment, to manipulate, position and deform 3D objects. First, we study the possibilities and limitations of the hand and finger gestures while interacting on a touch screen in order to discover which gestures are the most adapted to edit 3D scene and environment. In particular, we evaluate the effective number of degrees of freedom of the human hand when constrained on a planar surface. Meanwhile, we develop a new gesture analysis method using phases to identify key motion of the hand and fingers in real time. These results, combined to several specific user-studies, lead to a gestural design pattern which handle not only navigation (camera positioning), but also object positioning, rotation and global scaling. Then, this pattern is extended to complex deformation (such as adding and deleting material, bending or twisting part of objects, using local control). Using these results, we are able to propose and evaluate a 3D world editing interface that handle a natural touch interaction, in which mode selection (i.e. navigation, object positioning or object deformation) and task selections is automatically processed by the system, relying on the gesture and the interaction context (without any menu or button). Finally, we extend this interface to integrate more complex deformations, adapting the garment transfer from a character to any other in order to process interactive deformation of the garment while the wearing character is deformed.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Deslocamentos de côndilos e ramos acompanhando a cirurgia de avanço mandibular foram descritos, mas a influência destas mudanças na instabilidade pós-cirúrgica ainda é pouco conhecida. Limitações na avaliação quantitativa de deslocamentos cirúrgicos pelos métodos tradicionais podem ser superadas através da tomografia cone beam e ferramentas de imagens tridimensionais. O objetivo deste trabalho foi avaliar as mudanças nas posições de côndilos, ramos e mento após cirurgia isolada de avanço de mandíbula. Foram realizadas tomografias computadorizadas de feixe cônico (CBCT) antes da cirurgia (T1), 1 semana (T2) e 6 semanas após a cirurgia (T3) em 20 pacientes retrognatas com padrão normal ou horizontal (14 mulheres e 6 homens, idade média = 27,08 anos, variação de 13,9 a 53,7 anos) submetidos à osteotomia bilateral sagital do ramo para avanço mandibular, sete deles com mentoplastia como procedimento auxiliar. Modelos 3D computadorizados foram construídos e superpostos utilizando-se a base do crânio de T1 como referência. Para a superposição, utilizou-se um método automático que identifica e compara a escala de cinza dos voxels de duas estruturas tridimensionais, eliminando a necessidade de marcação de pontos anatômicos pelo operador. Após a combinação dos modelos das três fases, regiões anatômicas de interesse foram selecionadas, sendo analisadas separadamente. Distâncias entre as superfícies anatômicas foram computadas, no mesmo indivíduo, entre T1-T2, T2-T3 e T1-T3, exibindo as mudanças ocorridas com a cirurgia, após a remoção do splint cirúrgico, e a resultante total dos deslocamentos cirúrgicos, respectivamente. A navegação pelos modelos superpostos através dos métodos de visualização por mapas coloridos e semi-transparências permitiu a avaliação das direções de deslocamento das estruturas. Um deslocamento ântero-inferior do mento foi observado em todos os casos entre T1 e T2 (>4mm em 87,5%); entre T2 e T3 observou-se um movimento ântero-superior em 69% dos pacientes, e com algum componente posterior em 25% (<3mm). Entre T1-T3, observou-se um deslocamento ântero-inferior em 87,5% dos casos, e apenas anterior em 12,5% (>4mm em 80%). Considerando-se todas as direções de deslocamento, os côndilos apresentaram um movimento menor que 2mm em 77,5% (T1-T2) e 90% (T2-T3 e T1-T3) dos casos, enquanto os ramos deslocaram-se menos de 3mm em 72,5% (T1-T2) e menos de 2mm em 87,5% (T2-T3) e 82% (T1-T3) dos casos. Os côndilos apresentaram uma tendência de deslocamento látero-posterior com a cirurgia, e uma tendência oposta entre T2 e T3. Os ramos apresentaram tendência de deslocamento ínfero-latero-posterior com a cirurgia, mas súpero-medio-anterior entre T2 e T3. Entre T1-T3, as tendências foram médio-postero-superior nos côndilos e súpero-latero-anterior nos ramos. Importantes deslocamentos foram observados nos ramos e côndilos após o procedimento cirúrgico, mas as mudanças após a remoção do splint cirúrgico sugerem uma resposta adaptativa tendendo às posições pré-cirúrgicas. As mudanças no mento após seis semanas sugeriram adaptações aceitáveis na maioria dos casos. / Condyles and rami displacements following mandibular advancement surgery have been described, but these changes and bone remodeling/resorption influence on post-treatment instability are poorly understood at present. Cone beam computed tomography (CBCT) and 3D imaging tools can now overcome conventional methods limitations in assessing surgical displacements. The aim of this study was to assess surgery and short-term post-surgery changes in the position of the condyles, rami and chin with the superimposition of 3D cone-beam CT models. Pre-surgery (T1), 1 week post-surgery (T2), and 6 week post-surgery (T3) CBCT scans were acquired for 20 retrognathic patients (14 females and 6 males, mean age = 27.08 years, range = 13.9-53.7 years) with short or normal face height who underwent a BSSO mandibular advancement. Seven patients also received a genioplasty as an adjunctive procedure. 3D models were built and superimposed through a fully automated voxel-wise method using the cranial base of the pre-surgery scan as reference. After combining the 3 models, anatomic regions of the chin, right and left condyle and ramus were selected and analyzed separately. Within-subject surface distances between T1-T2, T2-T3, and T1-T3 were computed, showing changes with surgery, after splint removal, and the total movement resultant, respectively. Color-coded maps and semi-transparent display of overlaid structures allowed the evaluation of displacement directions. After an antero-inferior chin displacement with surgery in all the cases (>4mm in 87.5%), 25% of the patients showed some kind of posterior movement (< 3mm), and 69% showed an antero-superior movement after splint removal. Comparing T1-T3, an antero-inferior (87.5% of the cases) or only inferior (12.5%) displacement was observed (>4mm in 80%). Considering all directions of displacement, the surface distance differences for the condyles and rami were small: 77.5% of the condyles moved <2mm with surgery (T1-T2), and 90% moved <2mm in the short-term (T2-T3) and in the total evaluation (T1-T3), while the rami showed a <3mm change with surgery in 72.5% of the cases, and a <2mm change in 87.5% (T2-T3) and in 82% (T1-T3). The condyles tended to move latero-posteriorly with surgery compared to a medio-anterior movement between T2-T3. The rami exhibited outward movements with surgery (infero-latero-posterior) and supero-medio-anterior displacements from T2-T3. The displacement tendencies between T1-T3 were medio-postero-superior in the condyles, and supero-latero-anterior in the rami. Expected displacements with surgery were observed and post-surgery changes suggested a short-term adaptive response toward recovery of condyle and ramus displacements. The changes on the chin following splint removal suggested an acceptable adaptation.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Deslocamentos de côndilos e ramos acompanhando a cirurgia de avanço mandibular foram descritos, mas a influência destas mudanças na instabilidade pós-cirúrgica ainda é pouco conhecida. Limitações na avaliação quantitativa de deslocamentos cirúrgicos pelos métodos tradicionais podem ser superadas através da tomografia cone beam e ferramentas de imagens tridimensionais. O objetivo deste trabalho foi avaliar as mudanças nas posições de côndilos, ramos e mento após cirurgia isolada de avanço de mandíbula. Foram realizadas tomografias computadorizadas de feixe cônico (CBCT) antes da cirurgia (T1), 1 semana (T2) e 6 semanas após a cirurgia (T3) em 20 pacientes retrognatas com padrão normal ou horizontal (14 mulheres e 6 homens, idade média = 27,08 anos, variação de 13,9 a 53,7 anos) submetidos à osteotomia bilateral sagital do ramo para avanço mandibular, sete deles com mentoplastia como procedimento auxiliar. Modelos 3D computadorizados foram construídos e superpostos utilizando-se a base do crânio de T1 como referência. Para a superposição, utilizou-se um método automático que identifica e compara a escala de cinza dos voxels de duas estruturas tridimensionais, eliminando a necessidade de marcação de pontos anatômicos pelo operador. Após a combinação dos modelos das três fases, regiões anatômicas de interesse foram selecionadas, sendo analisadas separadamente. Distâncias entre as superfícies anatômicas foram computadas, no mesmo indivíduo, entre T1-T2, T2-T3 e T1-T3, exibindo as mudanças ocorridas com a cirurgia, após a remoção do splint cirúrgico, e a resultante total dos deslocamentos cirúrgicos, respectivamente. A navegação pelos modelos superpostos através dos métodos de visualização por mapas coloridos e semi-transparências permitiu a avaliação das direções de deslocamento das estruturas. Um deslocamento ântero-inferior do mento foi observado em todos os casos entre T1 e T2 (>4mm em 87,5%); entre T2 e T3 observou-se um movimento ântero-superior em 69% dos pacientes, e com algum componente posterior em 25% (<3mm). Entre T1-T3, observou-se um deslocamento ântero-inferior em 87,5% dos casos, e apenas anterior em 12,5% (>4mm em 80%). Considerando-se todas as direções de deslocamento, os côndilos apresentaram um movimento menor que 2mm em 77,5% (T1-T2) e 90% (T2-T3 e T1-T3) dos casos, enquanto os ramos deslocaram-se menos de 3mm em 72,5% (T1-T2) e menos de 2mm em 87,5% (T2-T3) e 82% (T1-T3) dos casos. Os côndilos apresentaram uma tendência de deslocamento látero-posterior com a cirurgia, e uma tendência oposta entre T2 e T3. Os ramos apresentaram tendência de deslocamento ínfero-latero-posterior com a cirurgia, mas súpero-medio-anterior entre T2 e T3. Entre T1-T3, as tendências foram médio-postero-superior nos côndilos e súpero-latero-anterior nos ramos. Importantes deslocamentos foram observados nos ramos e côndilos após o procedimento cirúrgico, mas as mudanças após a remoção do splint cirúrgico sugerem uma resposta adaptativa tendendo às posições pré-cirúrgicas. As mudanças no mento após seis semanas sugeriram adaptações aceitáveis na maioria dos casos. / Condyles and rami displacements following mandibular advancement surgery have been described, but these changes and bone remodeling/resorption influence on post-treatment instability are poorly understood at present. Cone beam computed tomography (CBCT) and 3D imaging tools can now overcome conventional methods limitations in assessing surgical displacements. The aim of this study was to assess surgery and short-term post-surgery changes in the position of the condyles, rami and chin with the superimposition of 3D cone-beam CT models. Pre-surgery (T1), 1 week post-surgery (T2), and 6 week post-surgery (T3) CBCT scans were acquired for 20 retrognathic patients (14 females and 6 males, mean age = 27.08 years, range = 13.9-53.7 years) with short or normal face height who underwent a BSSO mandibular advancement. Seven patients also received a genioplasty as an adjunctive procedure. 3D models were built and superimposed through a fully automated voxel-wise method using the cranial base of the pre-surgery scan as reference. After combining the 3 models, anatomic regions of the chin, right and left condyle and ramus were selected and analyzed separately. Within-subject surface distances between T1-T2, T2-T3, and T1-T3 were computed, showing changes with surgery, after splint removal, and the total movement resultant, respectively. Color-coded maps and semi-transparent display of overlaid structures allowed the evaluation of displacement directions. After an antero-inferior chin displacement with surgery in all the cases (>4mm in 87.5%), 25% of the patients showed some kind of posterior movement (< 3mm), and 69% showed an antero-superior movement after splint removal. Comparing T1-T3, an antero-inferior (87.5% of the cases) or only inferior (12.5%) displacement was observed (>4mm in 80%). Considering all directions of displacement, the surface distance differences for the condyles and rami were small: 77.5% of the condyles moved <2mm with surgery (T1-T2), and 90% moved <2mm in the short-term (T2-T3) and in the total evaluation (T1-T3), while the rami showed a <3mm change with surgery in 72.5% of the cases, and a <2mm change in 87.5% (T2-T3) and in 82% (T1-T3). The condyles tended to move latero-posteriorly with surgery compared to a medio-anterior movement between T2-T3. The rami exhibited outward movements with surgery (infero-latero-posterior) and supero-medio-anterior displacements from T2-T3. The displacement tendencies between T1-T3 were medio-postero-superior in the condyles, and supero-latero-anterior in the rami. Expected displacements with surgery were observed and post-surgery changes suggested a short-term adaptive response toward recovery of condyle and ramus displacements. The changes on the chin following splint removal suggested an acceptable adaptation.

Orientador: Jorge Stolfi / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-23T12:30:57Z (GMT). No. of bitstreams: 1 Pereira_DanilloRoberto_D.pdf: 2771811 bytes, checksum: 6d5b092c08b5c011636be5fc2661e4a0 (MD5) Previous issue date: 2013 / Resumo: Nesta tese descrevemos um algoritmo genérico (que denominamos MSFit) capaz de estimar a pose e as deformações de modelos 3D de estruturas biológicas (bactérias, células e etc.) em imagens obtidas por meio de microscópios óticos ou de varredura eletrônica. O algoritmo usa comparação multi-escala de imagens utilizando uma métrica sensível ao contorno; e um método original de otimização não-linear. Nos nossos testes com modelos de complexidade moderada (até 12 parâmetros) o algoritmo identifica corretamente os parâmetros do modelo em 60-70% dos casos com imagens reais e entre 80-90% dos casos com imagens sintéticas / Abstract: In this thesis we describe a generic algorithm (which we call MSFit) able to estimate the pose and deformations of 3D models of biological structures (bacteria, cells, etc.) with images obtained by optical and scanning electron microscopes. The algorithm uses an image comparison metric multi-scale, that is outline-sensitive, and a novel nonlinear optimization method. In our tests with models of moderate complexity (up to 12 parameters) the algorithm correctly identifies the model parameters in 60-70 % of the cases with real images and 80-90 % of the cases with synthetic images / Doutorado / Ciência da Computação / Doutor em Ciência da Computação

An exploratory research project was conducted through a company focusing on CAD and their own developed real-time 3D model viewer. The company needed to be able to convert CAD models to use in Unreal Engine with great visual quality. Before this project, another was conducted to perform the simple conversion of CAD models to the FBX file format, which Unreal uses. In extension to the previous project, one needed to add functionalities to manipulate the models for better quality and performance. The tasks were carried out and performed with good results.

Les travaux développés dans cette thèse s’intéressent à l’analyse fine des véhicules à partir d’une image. Nous définissons le terme d’analyse fine comme un regroupement des concepts suivants : la détection des véhicules dans l’image, l’estimation de leur point de vue (ou orientation), la caractérisation de leur visibilité, leur localisation 3D dans la scène et la reconnaissance de leur marque et de leur modèle. La construction de solutions fiables d’analyse fine de véhicules laisse place à de nombreuses applications notamment dans le domaine du transport intelligent et de la vidéo surveillance.Dans ces travaux, nous proposons plusieurs contributions permettant de traiter partiellement ou complètement cette problématique. Les approches mises en oeuvre se basent sur l’utilisation conjointe de l’apprentissage profond et de modèles 3D de véhicule. Dans une première partie, nous traitons le problème de reconnaissance de marques et modèles en prenant en compte la difficulté de la création de bases d’apprentissage. Dans une seconde partie, nous investiguons une méthode de détection et d’estimation du point de vue précis en nous basant sur l’extraction de caractéristiques visuelles locales et de la cohérence géométrique. La méthode utilise des modèles mathématiques uniquement appris sur des données synthétiques. Enfin, dans une troisième partie, un système complet d’analyse fine de véhicules dans le contexte de la conduite autonome est proposé. Celui-ci se base sur le concept d’apprentissage profond multi-tâches. Des résultats quantitatifs et qualitatifs sont présentés tout au long de ce manuscrit. Sur certains aspects de l’analyse fine de véhicules à partir d’une image, ces recherches nous ont permis de dépasser l’état de l’art. / In this thesis, we are interested in fine-grained analysis of vehicle from an image. We define fine-grained analysis as the following concepts : vehicle detection in the image, vehicle viewpoint (or orientation) estimation, vehicle visibility characterization, vehicle 3D localization and make and model recognition. The design of reliable solutions for fine-grained analysis of vehicle open the door to multiple applications in particular for intelligent transport systems as well as video surveillance systems. In this work, we propose several contributions allowing to address partially or wholly this issue. Proposed approaches are based on joint deep learning technologies and 3D models. In a first section, we deal with make and model classification keeping in mind the difficulty to create training data. In a second section, we investigate a novel method for both vehicle detection and fine-grained viewpoint estimation based on local apparence features and geometric spatial coherence. It uses models learned only on synthetic data. Finally, in a third section, a complete system for fine-grained analysis is proposed. It is based on the multi-task concept. Throughout this report, we provide quantitative and qualitative results. On several aspects related to vehicle fine-grained analysis, this work allowed to outperform state of the art methods.

Cette thèse présente le modèle 3D du bassin de la Plaine du Pô en Italie du nord. Les six parties de la thèse conduisent le lecteur à partir du cadre géologique de base aux géométries et à la cinématique de la déformation à travers la région, ainsi qu’aux possibles applications en milieu académique ou industriel. Le modèle a intégré des données éparses et de qualité inégale, tirées exclusivement de la littérature publique. L'ensemble de données utilisées pour la création du modèle se base strictement sur des données en profondeur (i.e. dans leur dimension de profondeur). Les données sismiques disponibles ont été intentionnellement écartées pour les raisons suivantes: a) elles sont mal distribuées à travers le secteur d'étude, b) elles se rapportent à des images de basse qualité, d) leur intégration dans le modèle aurait impliqué un long et difficile travail d’évaluation du meilleur modèle de vitesse de propagation des ondes sismiques dans les sédiments, le mieux à même d’être employé pour une conversion finale temps-profondeur, la variation latérale et verticale des vitesses sismiques à l’échelle régionale étant douteuse ou, au mieux, incertaine. La méthodologie appliquée, la création de modèles et l'analyse des améliorations du modèle 3D fournissent un certain nombre de conclusions sur la géométrie, le style structural et la cinématique de la Plaine du Pô et leur contribution en terme de sismicité du bassin et de son potentiel pétrolier, avec une confirmation mutuelle des, mais aussi par, les résultats locaux et épars obtenus par d’autres auteurs. Le résultat principal du projet est d'avoir prouvé la capacité du modèle à visualiser et analyser la complexité du bassin de la Plaine du Pô dans les 3 dimensions à différentes échelles d'observation, de l’échelle crustal jusqu'au niveau des prospects... / This thesis deals with the 3D model building of the Po Valley foreland basin in northern Italy. The six parts of the thesis lead from the basic geological framework to the deformation geometries and kinematics across the region, to some of the possible model applications, for both academia and industry. The model has integrated sparse and variable quality data, exclusively taken from the public literature. The complete dataset used for the performed model building, strictly relies on depth-data (i.e. in their depth dimension). As such, the few available seismic data have been intentionally left apart because: a) they are poorly distributed across the study-area, b) they definitely refer to low quality images, d) their integration into the model would have implied a long and difficult work about the definition of the most-likely sediment velocities to be used for an ultimate time-depth conversion, uncertain and, at best, questionable. The applied methodology, the related model building and the progressing analysis of 3D model results suggest and discuss a number of conclusions about the Po Valley structural geometries-style-kinematics. From such results can be derived implications on basin seismicity and hydrocarbon potential, while confirming (thus being supported by) the local and sparse results of previous authors. The major result from the project is to have proven the model capability in rendering and analyzing the entire Po Valley basin structural complexity in 3D dimensions, from crustal to field scale. Thanks to this, the model is unique in the literature of the region...

The representation of 3D polygonal model on several levels of available detail is a problem inherent in the process of rendering a scene. Highly-detailed models, if placed far from the camera, suffer from spatial aliasing that results from inadequate sampling of their surface, and require disproportionately large amount of time to render. Low-detailed models on the other hand reduce the visual quality of the scene when placed too near to the camera. This report delves in both the theory and the practical techniques used for solving these problems. It describes various published solutions and the principles behind them, and presents a design and an implementation of selected techniques for the GPUEngine library.