Browsing by Author "Vanhoey, Kenneth"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    A Halfedge Refinement Rule for Parallel Catmull-Clark Subdivision
    (The Eurographics Association and John Wiley & Sons Ltd., 2021) Dupuy, Jonathan; Vanhoey, Kenneth; Binder, Nikolaus and Ritschel, Tobias
    We show that Catmull-Clark subdivision induces an invariant one-to-four refinement rule for halfedges that reduces to simple algebraic expressions. This has two important consequences. First, it allows to refine the halfedges of the input mesh, which completely describe its topology, concurrently in breadth-first order. Second, it makes the computation of the vertex points straightforward as the halfedges provide all the information that is needed. We leverage these results to derive a novel parallel implementation of Catmull-Clark subdivision suitable for the GPU. Our implementation supports non-quad faces, extraordinary vertices, boundaries and semi-sharp creases seamlessly. Moreover, we show that its speed scales linearly with the number of processors, and yields state-of-the-art performances on modern GPUs.
  • Thumbnail Image
    Item
    A Halfedge Refinement Rule for Parallel Loop Subdivision
    (The Eurographics Association, 2022) Vanhoey, Kenneth; Dupuy, Jonathan; Pelechano, Nuria; Vanderhaeghe, David
    We observe that a Loop refinement step invariably splits halfedges into four new ones. We leverage this observation to formulate a breadth-first uniform Loop subdivision algorithm: Our algorithm iterates over halfedges to both generate the refined topological information and scatter contributions to the refined vertex points. Thanks to this formulation we limit concurrent data access, enabling straightforward and efficient parallelization on the GPU. We provide an open-source GPU implementation that runs at state-of-the-art performances and supports production-ready assets, including borders and semi-sharp creases.
  • No Thumbnail Available
    Item
    Unsupervised Deep Single-Image Intrinsic Decomposition using Illumination-Varying Image Sequences
    (The Eurographics Association and John Wiley & Sons Ltd., 2018) Lettry, Louis; Vanhoey, Kenneth; Van Gool, Luc; Fu, Hongbo and Ghosh, Abhijeet and Kopf, Johannes
    Machine learning based Single Image Intrinsic Decomposition (SIID) methods decompose a captured scene into its albedo and shading images by using the knowledge of a large set of known and realistic ground truth decompositions. Collecting and annotating such a dataset is an approach that cannot scale to sufficient variety and realism. We free ourselves from this limitation by training on unannotated images. Our method leverages the observation that two images of the same scene but with different lighting provide useful information on their intrinsic properties: by definition, albedo is invariant to lighting conditions, and cross-combining the estimated albedo of a first image with the estimated shading of a second one should lead back to the second one's input image. We transcribe this relationship into a siamese training scheme for a deep convolutional neural network that decomposes a single image into albedo and shading. The siamese setting allows us to introduce a new loss function including such cross-combinations, and to train solely on (time-lapse) images, discarding the need for any ground truth annotations. As a result, our method has the good properties of i) taking advantage of the time-varying information of image sequences in the (pre-computed) training step, ii) not requiring ground truth data to train on, and iii) being able to decompose single images of unseen scenes at runtime. To demonstrate and evaluate our work, we additionally propose a new rendered dataset containing illumination-varying scenes and a set of quantitative metrics to evaluate SIID algorithms. Despite its unsupervised nature, our results compete with state of the art methods, including supervised and non data-driven methods.