SAC-GAN: Structure-Aware Image Composition.

We introduce an end-to-end learning framework for image-to-image composition, aiming to plausibly compose an object represented as a cropped patch from an object image into a background scene image. As our approach emphasizes more on semantic and structural coherence of the composed images, rather than their pixel-level RGB accuracies, we tailor the input and output of our network with structure-aware features and design our network losses accordingly, with ground truth established in a self-supervised setting through the object cropping. Specifically, our network takes the semantic layout features from the input scene image, features encoded from the edges and silhouette in the input object patch, as well as a latent code as inputs, and generates a 2D spatial affine transform defining the translation and scaling of the object patch. The learned parameters are further fed into a differentiable spatial transformer network to transform the object patch into the target image, where our model is trained adversarially using an affine transform discriminator and a layout discriminator. We evaluate our network, coined SAC-GAN, for various image composition scenarios in terms of quality, composability, and generalizability of the composite images. Comparisons are made to state-of-the-art alternatives, including Instance Insertion, ST-GAN, CompGAN and PlaceNet, confirming superiority of our method.

RIAS: Repeated Invertible Averaging for Surface Multiresolution of Arbitrary Degree.

In this article, we introduce two local surface averaging operators with local inverses and use them to devise a method for surface multiresolution (subdivision and reverse subdivision) of arbitrary degree. Similar to previous works by Stam, Zorin, and Schröder that achieved forward subdivision only, our averaging operators involve only direct neighbours of a vertex, and can be configured to generalize B-Spline multiresolution to arbitrary topology surfaces. Our subdivision surfaces are hence able to exhibit Cd continuity at regular vertices (for arbitrary values of d) and appear to exhibit C1 continuity at extraordinary vertices. Smooth reverse and non-uniform subdivisions are additionally supported.

Physical Visualization of Geospatial Datasets.

Geospatial datasets are too complex to easily visualize and understand on a computer screen. Combining digital fabrication with a discrete global grid system (DGGS) can produce physical models of the Earth for visualizing multiresolution geospatial datasets. This proposed approach includes a mechanism for attaching a set of 3D printed segments to produce a scalable model of the Earth. The authors have produced two models that support the attachment of different datasets both in 2D and 3D format.