It is commonly believed that image recognition based on RGB improves when using RGB-D, ie: when depth information (distance from the camera) is added. Adding depth should make models more robust to appearance variations in colors and lighting; to recognize shape and spatial relationships while allowing models to ignore irrelevant backgrounds. In this paper we investigate how robust current RGB-D models truly are to changes in appearance, depth, and background where we vary one modality (RGB or depth) and compare RGB-D to RGB-only and depth-only in a semantic segmentation setting. Experiments show that all investigated RGB-D models show some robustness to variations in color, but might severely fail for unseen variations in lighting, spatial position and backgrounds. Our results show that we need new RGB-D models that can exploit the best of both modalities while remaining robust to changes in a single modality.

Dimensionality reduction is essential for analyzing high-dimensional datasets across various fields. While t-SNE is a popular method for this purpose in Euclidean spaces, recent advancements suggest that hyperbolic spaces can better represent hierarchical structures. However, there is a notable lack of data structures and algorithms tailored for hyperbolic spaces. This research addresses this gap by implementing a hyperbolic quadtree structure in the upper half-plane model and integrating it into the hyperbolic t-SNE framework. Our goal is to accelerate the optimization of the hyperbolic t-SNE while maintaining reasonable precision and recall. We conduct rigorous benchmarking experiments to evaluate the performance of this approach, comparing it to existing methods. The findings provide insights into the practical utility of using the hyperbolic quadtree structure in the upper half-plane model in hyperbolic t-SNE embeddings.

Recovering the appearance and physical parameters of elastic objects from multi-view video is essential for many applications that require simulation of the real world. Past methods for this task have provided accurate results in recovering physical properties; however, their re- liance on Neural Radiance Fields (NeRFs) for novel view synthesis means they trade off visual quality for rendering speed. To address this issue, we present a novel framework for the joint appearance reconstruction and physical parameter estimation of elastic objects relying on 3D Gaussian splatting. Our key insight is that dynamic 3D Gaussian kernels extracted from multi-view video can be used to reconstruct the object’s geometry and supervise elastic parame- ter fitting through a differentiable physics engine. Novel views and object behaviours can then be constructed by forward simulating the extracted mesh and using it to drive the Gaussian kernels. We demon- strate that our method is competitive with the state of the art on phys- ical parameter estimation while being better at reconstructing object appearance. Additionally, our method can simulate novel views and object interactions at near real-time rates that outperform past ap- proaches.

The n-body problem is the simulation of pair-wise interactions between n objects. This problem appears in many forms, with the classic example being the modeling of gravitational forces between point masses, necessary for cosmological simulations. Many approximation approaches have been devised to reduce the complexity of this problem.

t-SNE is a data visualization method that requires repeatedly solving a variant of the n-body problem. A recent paper (An Efficient Dual-Hierarchy t-SNE Minimization, van de Ruit et. al.) proposes a novel algorithm that outperforms other t-SNE minimization methods on medium-scale datasets. The report proves the viability of a dual-traversal method that uses an embedding tree to emit forces and an independent field tree to collect forces. Because the embedding tree is a Linear-BVH and the field tree is an orthtree built to a fixed depth, the overall algorithm has linear complexity.

This thesis demonstrates how the dual-tree approach can be adapted for gravitational n-body simulations. Following this, it measures the performance against similar implementations of other algorithms and shows that while the adapted Dual Hierarchy approach is faster than Barnes-Hut, it is outperformed by the Fast Multipole Method on realistic large-scale cosmological datasets.

Helpline counselors can be trained faster using agent-based social skills training systems. These systems utilize conversational agents that simulate interactions with users and provide feedback. This research evaluates the noticeability of behavior changes in such a conversational agent. By examining participants' ability to detect and interpret these changes, valuable insights can be gained to enhance the agent's representation of real-life scenarios and improve its effectiveness as a training tool.

A mixed-methods approach was employed, combining quantitative data using descriptive statistics and qualitative data through content analysis. The quantitative analysis revealed a precision score of 93% and a recall score of 36%, indicating a higher accuracy in correctly identifying behavior changes compared to capturing all instances of change. The qualitative analysis revealed that participants' perceptions fell within four distinct categories: communication style, positive emotion, negative emotion, and tone & attitude. Participants demonstrated heightened attentiveness to changes in positive behavior, particularly instances of happiness expressed by the agent, and displayed less attention towards changes in negative behavior.

In conclusion, this research demonstrates that behavior changes in the conversational agent are noticeable, albeit with varying degrees of attention across different types of behavior. Further exploration with a larger sample size is warranted to ascertain the extent of noticeability. The diverse range of behavior changes indicates the agent's capability to adapt and exhibit visible shifts in behavior. These insights can inform the refinement and development of agent-based training systems, ultimately enhancing the training experiences of chat-based helpline counselors.

Chatbots are tools that can potentially be utilized in chat-based child helpline training. In this type of training, the quality of the feedback received is of vital importance. This paper aims to analyze the automated feedback generated by such a chatbot. The domains analyzed include user comprehension, usefulness, and potential improvement points. In a user study, a formative assessment and two interviews were conducted for each domain, respectively.
For comprehensibility, all participants could easily understand the feedback report. They found that the bot could be easier to work around after reading the feedback, with the table being of much guidance. They found the transcript to be a welcome addition, but missing constructive feedback. Regarding improvement points, two of them were tightly related to the limitations of the chatbot, rather than the report itself. Extra guidance and instructions were deemed necessary by the participants, alongside an easier-to-read transcript interface.

The aim of this research is to evaluate the believability of Lilobot, a conversational agent meant to act as a virtual child for training helpline workers. Numerous aspects of believability are explored by means of a user study involving a questionnaire and interview with 10 participants. Questionnaire results indicate that improvement to the chatbot's believability is likely necessary.
The findings from the interviews are that the use of emoticons and acknowledging the context of the application raise believability, while unresponsiveness and repeated utterances lower it. While Lilobot did express valid and real emotions, study participants suggested improving the appropriateness of its reactions and expanding its vocabulary.

Agent-based social skills training systems have been gaining attention for their potential to improve social skills development in various contexts. Through a rapid review methodology, data was collected from diverse sources, including company websites and research papers. This study then uses the collected data to categorize 8 commercial systems based on their agent model and feedback approaches, into two categorization tables. The findings reveal notable trends in the use of choice-based input, scenario-defined decision-making, and post-interaction feedback. Additionally, the paper discusses the limitations of these findings, highlights characteristics of commercial systems and compares them to research systems, as well as suggesting areas for future research. This study contributes to the understanding and advancement of agent-based social skills training systems, offering guidance to researchers in this field.

During the preoperative planning for breast-conserving surgery, the surgeon makes use of an MRI scan of the breast cancer patient in the prone position to accurately locate the tumour. However, surgery is performed with the patient in the supine position. The surgeon needs to mentally translate the location of the tumour from the prone position towards the supine position. The usage of mixed reality systems in the form of the Microsoft HoloLens could visually aid the surgeon by projecting the tumour in the supine position onto the patient. This requires us to obtain the tumour in the supine position by dealing with the prone-to-supine breast image registration problem.
We propose to maximize the overlap between the prone and supine surfaces of the breast through the use of soft constraints to guide the deformation. An initial subspace FEM simulation pulls the prone and supine surfaces towards each other at specified landmark correspondences between the surfaces. The resulting prone surface is projected onto the supine surface and these surface vertices are restricted to tangential movement across the surface of the breast. The tissues at the interior of the breast are deformed accordingly. Our experiments with synthesized data indicate the robustness of our method under various scenarios, except for the rigid misalignment between the prone and supine meshes. This problem of misalignment also affects the experiments with real-world data, which results in distances greater than 12mmbetween the centroids of the resulting deformed tumour and the ground-truth tumour.

The goal of this project was to simulate the colour change over time of Girl with a Pearl Earring from Johannes Vermeer. Using art history knowledge and computational methods this was proven to be possible. The combination of these methods is unique and promising.

As a first for the Girl, the paints that Vermeer used have been reconstructed and artificially aged by UV light at AkzoNobel. Their colour change was then measured and turned out to be as expected. The mechanisms of indigo, yellow and red lake fading seems plausible. A start in reconstructing the painting has been made by digitally applying these colour changes on the original painting.

The digital reconstruction showed that the contrast increased and the Girl is more lively and looks ready to step out of the painting, there is a clear three dimensional effect. The colours are more defined instead of faded and patchy. More contrast can be seen, which is for example shown in the white collar. It also makes the edges between the face and background sharper. Overall, the painting shows more Houding and seems to be more like Vermeer probably intended it to be. A reconstruction made with BRDF should be more scientifically accurate than the reconstruction presented here.

Important points for future research are more opaque samples, which are created in more batches. Another point is that the measured light intensities should not be in lux and to be quick in the BRDF measurements after UV aging. Finally, it would be interesting to reconstruct other paintings, from Vermeer or from another painter.

The Hierarchical Subspace Iteration Method is a novel method used to compute eigenpairs of the Laplace-Beltrami problem. It reduces the number of iterations required for convergence by restricting the problem to a smaller space and prolonging the solution as a starting point. This method has shown great performance improvements for Laplace-Beltrami eigenproblems.
We propose an adaptation to the Hierarchical Subspace Iteration Method that allows for computing vibration modes of elastic objects. We evaluate potential optimizations that can be made, as well as the performance characteristics of the method. Our method was shown to be faster than SIM in most cases while even beating Matlab's Lanczos solver in some cases.

Virtual worlds are rapidly increasing in size, enabled by advancements in computing technology, which puts a large burden on designers creating them.
Procedural Content Generation can help alleviate this burden, though lacks precise control, to the detriment of designer intent. Some PCG algorithms, such as Wave Function Collapse (WFC), known for generating tile-based content adhering to certain constraints, are able to induce this control through mixed-initiative editing opportunities, combining the efforts of humans and machines while capturing designer intent. However, stock WFC operates on a flat set of tiles with all their semantics blurred together, which unnecessarily strains designers with additional cognitive load when working with such detailed tiles.



We therefore propose Hierarchical Semantic Wave Function Collapse (HSWFC), a generalized approach to WFC that augments the tileset with a new type of tile, the meta-tile, which represents semantic traits, and then organizes the tileset into a hierarchy akin to a taxonomy induced by such semantic representations: the meta-tree. A cell once collapsed to any tile high up in the meta-tree (such as "forest" or "village"), can further collapse to concrete tiles at the bottom (such as “tree” or “wall”).


We investigate how this extension in data organization affects the original algorithm, and explore several novel editing facilities, including e.g. sketching with semantic tiles, controlling tile distributions, regenerating areas that represent specific semantics, and more. A prototypical HSWFC-driven tile-editor was developed and evaluated through a user study, confirming that such an editor indeed reduces cognitive load compared to its stock WFC counterpart, and that the newly enabled features are highly valued by environment designers.

In this work, we present VisuaLayered, the implementation of a combined analysis workflow for pigment identification. VisuaLayered is an integrated, interactive system that focuses on the combined visual analysis of Macro X-Ray Fluorescence (MA-XRF) and Reflectance Imaging Spectroscopy (RIS) data.

Analysing paintings for pigment identification is relevant for many applications in the cultural heritage domain, such as conservation and restoration.
Domain experts use non-invasive scanning techniques as an initial step in their analysis. Two such techniques are MA-XRF and RIS. They provide hyperspectral data on the elemental and molecular composition of pigments, respectively. Domain experts analyse these two complementary data modalities in order to determine the pigments present in the different paint layers of a painting. However, due to the size and high-dimensionality of these datasets, experts have problems with efficiently analysing the data. In general, they examine the two data modalities separately in the analysis workflow and use their knowledge to unify all the information without additional software support, as there is no integrated system that is designed specifically for the combined analysis of MA-XRF and RIS data.

We worked in collaboration with domain experts from the Rijksmuseum, Amsterdam in order to design and implement VisuaLayered based on current domain practices. We use t-SNE for projecting the high-dimensional data into a two dimensional space, which the user can interactively explore in combination with other linked views in order to find connections between the two data modalities. With our system, experts can explore the spatial distribution and the correlation between pixels that have similar molecular and/or elemental compositions. Additionally, for the RIS data, we support endmember identification and analysis based on the pigments' spectral profiles.
We tested the efficiency of our system with respect to the designed workflow in a case study evaluation with our collaborator. They successfully used VisuaLayered for the analysis of one painting and found the views combining the two data modalities very useful for better understanding the relation between them. Moreover, they were able to identify new pigments, that they missed when using existing software.

Grid-based fluid simulations are often limited in resolution by their high memory usage and computational costs. One approach to reducing memory usage and computational costs is to vary the grid resolution over the spatial domain. We introduce DCGrid, a new data structure for fluid simulations. DCGrid is suited for instantiation in GPU memory and allows for varying resolution over the spatial domain. We developed an efficient, optimization-based method for local mesh refinement that automatically adapts the grid resolution according to user-defined parameters during simulations. Additionally, we complement our data structure with an efficient scheme for approximate handling of collisions between fluid and static solids on cells with varying resolutions. We integrate DCGrid in a cloud simulation and extend a terrain-atmosphere interaction model to work with cells of varying resolution and rapidly changing conditions. Furthermore, we demonstrate the performance of our new methods on both simple simulations of smoke flow and complex simulations of weather phenomena and compare them to similar fluid simulations on state-of-the-art adaptive grid structures.

This research builds upon a previous method of vectorisation of pixel art by pixel neighbour connecting and boundary energy minimisation. The downside of the previous method is the lack of user input throughout the process and possible divergence between the results and the artist’s vision. The proposed method uses the first part of previous method to connect neighbour pixels and continues with a spring simulation of the boundaries. A process that can be heavily user-guided with adjusting the stiffness of the springs. Results show the proposed method to achieve higher variance in the results, possibly resulting in output images closer to the artist’s vision. However, introduce lengthy process for the user that can be removed by providing higher level GUI.

Minimalistic, low-poly images have gotten more popular in recent years. But making these low-poly images by hand can be time intensive. For this reason, a lot of research has gone into how to triangulate an image (semi-)automatically. In previous works, only constant color and bilinear interpolation have been used to color the triangles in the triangulation. In this paper, different coloring methods have been investigated and showcased on a fixed mesh. They show some promising first results, but these methods can still be improved upon. When that happens, these coloring methods, combined with previous works on how to optimally triangulate, may result in more accurate or visually pleasing low-poly images.

In this paper, a method is proposed to artificially expand the dynamic range of screens with a limited dynamic range. This research is linked to a new film-making technology where, instead of using a green screen, the background of a scene is displayed on a screen in real time using a computer generated background. This provides real time lighting in the studio; however, due to the limited dynamic range of the screen, it can not fully replicate the brightness of light sources. Overcoming this problem involves capturing and synchronize frames that each display a small section of the wider dynamic range, defined as illumination maps. The method uses a pipeline in which the illumination maps are displayed on a monitor in a grouped order, which are then captured with a camera. The recording is processed by labeling the frames and selecting key frames. The key frames are then additively combined with compatible illumination maps, which result in a video of the full dynamic range.
A program was developed as a proof of concept, providing expected results. For various recording inputs, It was also found that the implemented program discarded a lot of the frames of the recordings. A variation of the proposed method also yielded a slight speed-up, for practically the same results.
The proposed method provides a good starting point tackling the problem of artificially extending the dynamic range. The program used is a step in the right direction, but has flaws that limit its usefulness.

This paper accelerates the rendering of diffusion curves using the ray tracing cores found on modern NVIDIA graphics cards using the method described by Bowers et al. [Bowers, Leahey and Wang 2011]. This method approximates the final result of the Poisson equation and in this paper is accelerated using Optix and dedicated ray tracing hardware. Using this method yielded a render time decrease of around 8 times between the Quadro P1000 and RTX 2060, while retaining the complex colour gradients and ease of use of traditional diffusion curves solved using the Poisson equation.

Video-streaming services usually feature post-processing effects to replace the background. However, these often yield inconsistent lighting. Machine-learning-based relighting methods can address this problem, but, at real-time rates, are restricted to a low resolution and can result in an unrealistic skin appearance. Physically-based rendering requires suitable skin models, which require specialised equipment. Our method is lightweight and uses only a standard smartphone. By correcting imperfections during capture, we extract a convincing physically-based skin model. In combination with suitable acceleration techniques, we achieve real-time rates on commodity hardware.

Suppose we have a target radiance distribution, a light source, and a plane for receiving light. How do we design a reflector that can give a similar result as the target radiance distribution? The inverse reflector problem is of high interest for light designers and related industries, such as lamp manufacture, headlight and street light design, and interior designs. There have been some researches on this specific topic. Nevertheless, the existing algorithms are either not fully compatible with parallel acceleration or have a narrow application scope (can only handle the far-field problem). This thesis's goal is to design a method that can have a fast approximation of the inverse reflector problem and handle different application scenarios. The core of the proposed method in this thesis is a modified simulated annealing algorithm. I first give the definition of the inverse reflector problem and explain why I choose simulated annealing in the proposed method by analysing the related works and various optimisation algorithms. Then, I detail what issues the plain simulated annealing algorithm will have and why it cannot achieve satisfactory results. Moreover, to solve the issues and improve performance, I propose additional strategies, such as Phong tessellation, randomisation strategy, multi-level strategy, history-decision strategy and penalising strategy. I also discuss the system design based on the proposed method and how to accelerate the optimisation process in this system. I evaluate and validate the proposed method by running test cases in different scenarios and compare the results with other algorithms. The results show that the method, as a general optimisation algorithm, can achieve good results in different application scenarios. The compatibility, speed and accuracy are the main advantages. The result of the proposed method can also serve as the initial guess for a finer optimisation.