This dissertation develops intrinsic approaches to learning and computing on curved surfaces. Specifically, we work on three tasks: analyzing 3D shapes using convolutional neural networks (CNNs), solving linear systems on curved surfaces, and recovering appearance properties from curved surfaces using multi-view capture. We argue that we can find more efficient and better performing algorithms for these tasks by using intrinsic geometry.

Chapter two and three consider CNNs on curved surfaces. We would like to find patterns with meaningful directional information, such as edges or corners.
On images, it is straightforward to define a convolution operator that encodes directional information, as the pixel grid provides a global reference for directions. Such a global coordinate system is not available for curved surfaces. Chapter two presents Harmonic Surface Networks. We apply a 2D kernel to the surface by using local coordinate systems. These local coordinate systems could be rotated in any direction around the normal, which is a problem for consistent pattern recognition. We overcome this ambiguity by computing complex-valued, rotation-equivariant features and transporting these features between coordinate systems with parallel transport along shortest geodesics.

Chapter three presents DeltaConv. DeltaConv is a convolution operator based on geometric operators from vector calculus, such as the Laplacian. A benefit of the Laplacian is that it is invariant to local coordinate systems. This solves the problem of a missing global coordinate system. However, the Laplacian operator is also isotropic. That means it cannot pick up on directional information. DeltaConv constructs anisotropic operators by splitting the Laplacian into gradient and divergence and applying a non-linearity in between. The resulting convolution operators are demonstrated on learning tasks for point clouds and achieve state-of-the-art results with a relatively simple architecture.

Chapter four considers solving linear systems on curved surfaces. This is relevant for many applications in geometry processing: smoothing data, simulating or animating 3D shapes, or machine learning on surfaces. A common way to solve large systems on grid-based data is a multigrid method. Multigrid methods require a hierarchy of grids and the operators that map between the levels in the hierarchy. We show that these components can be defined for curved surfaces with irregularly spaced samples using a hierarchy of graph Voronoi diagrams. The resulting approach, Gravo Multigrid, achieves solving times comparable to the state-of-the-art, while taking an order of magnitude less time for pre-processing: from minutes to seconds for meshes with over a million vertices.

Chapter five demonstrates the use of intrinsic geometry in the setting of appearance modeling, specifically capturing spatially-varying bidirectional reflectance distribution functions (SVBRDF). A low-cost setup to recover SVBRDFs is to capture photographs from multiple viewpoints. A challenge here, is that some reflectance behavior only shows up under certain viewing positions and lighting conditions, which means that we might not be able to tell one material type from another. We frame this as a question of (un)certainty: how certain are we, based on the input data? We build on previous work that shows that the reflection function can be modeled as a convolution of the BRDF with the incoming light. We propose improvements to the convolution model and develop algorithms for uncertainty analysis fully contained in the frequency domain. The result is a fast and uncertainty-aware SVBRDF recovery on curved surfaces.@en

We introduce a geometric multigrid method for solving linear systems arising from variational problems on surfaces in geometry processing, Gravo MG. Our scheme uses point clouds as a reduced representation of the levels of the multigrid hierarchy to achieve a fast hierarchy construction and to extend the applicability of the method from triangle meshes to other surface representations like point clouds, nonmanifold meshes, and polygonal meshes. To build the prolongation operators, we associate each point of the hierarchy to a triangle constructed from points in the next coarser level. We obtain well-shaped candidate triangles by computing graph Voronoi diagrams centered around the coarse points and determining neighboring Voronoi cells. Our selection of triangles ensures that the connections of each point to points at adjacent coarser and finer levels are balanced in the tangential directions. As a result, we obtain sparse prolongation matrices with three entries per row and fast convergence of the solver. Code is available at https://graphics.tudelft.nl/gravo_mg.@en

Cleaning and provisional restoration treatments on The Crucifixion (1425) revealed that the visible azurite layer obscures an originally golden background. This leads to the question of whether or not the azurite should be removed as it is not original, or should be kept as part of the panel’s history. To overcome this dilemma, this paper presents a methodological solution of approaching this problem by combining the knowledge of art historians, restorers and by integrating modern 3D technologies before, during, and after the restoration. By creating a modifiable digital model, a virtual restoration can potentially attribute to analyzing and visualizing optical changes due to restoration treatments. Additionally, multiple 3D-printed facsimiles physicalize these adaptations. A facsimile’s value is demonstrated by analyzing the diversity of meanings an artwork can have in terms of authenticity whilst respecting the artwork's material and social integrity. The 3D prints be decisive in the restoration of the panel.@en

Deep learning has improved vanishing point detection in images. Yet, deep networks require expensive annotated datasets trained on costly hardware and do not generalize to even slightly different domains, and minor problem variants. Here, we address these issues by injecting deep vanishing point detection networks with prior knowledge. This prior knowledge no longer needs to be learned from data, saving valuable annotation efforts and compute, unlocking realistic few-sample scenarios, and reducing the impact of domain changes. Moreover, the interpretability of the priors allows to adapt deep networks to minor problem variations such as switching between Manhattan and non-Manhattan worlds. We seamlessly incorporate two geometric priors: (i) Hough Transform -- mapping image pixels to straight lines, and (ii) Gaussian sphere -- mapping lines to great circles whose intersections denote vanishing points. Experimentally, we ablate our choices and show comparable accuracy to existing models in the large-data setting. We validate our model's improved data efficiency, robustness to domain changes, adaptability to non-Manhattan settings.@en

Learning from 3D point-cloud data has rapidly gained momentum, motivated by the success of deep learning on images and the increased availability of 3D~data. In this paper, we aim to construct anisotropic convolution layers that work directly on the surface derived from a point cloud. This is challenging because of the lack of a global coordinate system for tangential directions on surfaces. We introduce DeltaConv, a convolution layer that combines geometric operators from vector calculus to enable the construction of anisotropic filters on point clouds. Because these operators are defined on scalar- and vector-fields, we separate the network into a scalar- and a vector-stream, which are connected by the operators. The vector stream enables the network to explicitly represent, evaluate, and process directional information. Our convolutions are robust and simple to implement and match or improve on state-of-the-art approaches on several benchmarks, while also speeding up training and inference.@en

Multimodal imaging is used by conservators and scientists to study the composition of paintings. To aid the combined analysis of these digitisations, such images must first be aligned. Rather than proposing a new domain-specific descriptor, we explore and evaluate how existing feature descriptors from related fields can improve the performance of feature-based painting digitisation registration. We benchmark these descriptors on pixel-precise, manually aligned digitisations of ''Girl with a Pearl Earring'' by Johannes Vermeer (c. 1665, Mauritshuis) and of ''18th-Century Portrait of a Woman''. As a baseline we compare against the well-established classical SIFT descriptor. We consider two recent descriptors: the handcrafted multimodal MFD descriptor, and the learned unimodal SuperPoint descriptor. Experiments show that SuperPoint starkly increases description matching accuracy by 40% for modalities with little modality-specific artefacts. Further, performing craquelure segmentation and using the MFD descriptor results in significant description matching accuracy improvements for modalities with many modalityspecific artefacts.@en

This paper is concerned with a fundamental problem in geometric deep learning that arises in the construction of convolutional neural networks on surfaces. Due to curvature, the transport of filter kernels on surfaces results in a rotational ambiguity, which prevents a uniform alignment of these kernels on the surface. We propose a network architecture for surfaces that consists of vector-valued, rotation-equivariant features. The equivariance property makes it possible to locally align features, which were computed in arbitrary coordinate systems, when aggregating features in a convolution layer. The resulting network is agnostic to the choices of coordinate systems for the tangent spaces on the surface. We implement our approach for triangle meshes. Based on circular harmonic functions, we introduce convolution filters for meshes that are rotation-equivariant at the discrete level. We evaluate the resulting networks on shape correspondence and shape classifications tasks and compare their performance to other approaches.@en

Many universities digitize exams or the process of grading the exams. This potentially allows for faster grading, is less labor intensive and less error-prone. But are the grades produced by online grading consistent with how we grade on paper? In this paper we present preliminary results of the comparison between scores given by grading online and grading on paper.

@en

During the Second World War the German occupants of the Netherlands made ample use of the Scheveningen prison near The Hague, popularly nicknamed the Oranjehotel. One former death cell in this infamous prison (Doodencel 601) has been preserved in its original condition, showing wartime inscriptions on the cell walls. Interestingly, a small section of the wall has been given an additional plaster layer, presumably covering inscriptions. Here, we report on the visualization of this enigmatic text, which so far had escaped the reach of historians. Our visualization methodology was threefold. First, we determined the cell-wall stratigraphy and its composition based on a sample cross-section. Second, we prepared a physical model wall, mimicking the layering of the original cell wall. Third, we tested a combination of raking light photography and infrared thermography on the model wall. Applying this methodology on the original wall revealed the inscriptions, including the author’s name Daniël de Blocq van Scheltinga, a prominent Nazi collaborator, as well as a calendar and an important date of his post-war trial in the fall of 1945. Our visualizations flawlessly dovetail with archival findings. Together, they offer an intimate view of an early post-war inmate of the Scheveningen prison, whose message was covered up once the cell was transformed into a war monument in 1946.

@en