Geronazzo M, Vieira LS, Nilsson NC, Udesen J, and Serafin S
IEEE transactions on visualization and computer graphics [IEEE Trans Vis Comput Graph] 2020 May; Vol. 26 (5), pp. 1912-1922. Date of Electronic Publication: 2020 Feb 13.
Directivity and gain in microphone array systems for hearing aids or hearable devices allow users to acoustically enhance the information of a source of interest. This source is usually positioned directly in front. This feature is called acoustic beamforming. The current study aimed to improve users' interactions with beamforming via a virtual prototyping approach in immersive virtual environments (VEs). Eighteen participants took part in experimental sessions composed of a calibration procedure and a selective auditory attention voice-pairing task. Eight concurrent speakers were placed in an anechoic environment in two virtual reality (VR) scenarios. The scenarios were a purely virtual scenario and a realistic 360° audio-visual recording. Participants were asked to find an individual optimal parameterization for three different virtual beamformers: (i) head-guided, (ii) eye gaze-guided, and (iii) a novel interaction technique called dual beamformer, where head-guided is combined with an additional hand-guided beamformer. None of the participants were able to complete the task without a virtual beamformer (i.e., in normal hearing condition) due to the high complexity introduced by the experimental design. However, participants were able to correctly pair all speakers using all three proposed interaction metaphors. Providing superhuman hearing abilities in the form of a dual acoustic beamformer guided by head and hand movements resulted in statistically significant improvements in terms of pairing time, suggesting the task-relevance of interacting with multiple points of interests.
IEEE transactions on pattern analysis and machine intelligence [IEEE Trans Pattern Anal Mach Intell] 2019 Jun; Vol. 41 (6), pp. 1486-1500. Date of Electronic Publication: 2018 May 15.
Recognising human attributes from surveillance footage is widely studied for attribute-based re-identification. However, most works assume coarse, expertly-defined categories, ineffective in describing challenging images. Such brittle representations are limited in descriminitive power and hamper the efficacy of learnt estimators. We aim to discover more relevant and precise subject descriptions, improving image retrieval and closing the semantic gap. Inspired by fine-grained and relative attributes, we introduce super-fine attributes, which now describe multiple, integral concepts of a single trait as multi-dimensional perceptual coordinates. Crowd prototyping facilitates efficient crowdsourcing of super-fine labels by pre-discovering salient perceptual concepts for prototype matching. We re-annotate gender, age and ethnicity traits from PETA, a highly diverse (19K instances, 8.7K identities) amalgamation of 10 re-id datasets including VIPER, CUHK and TownCentre. Employing joint attribute regression with the ResNet-152 CNN, we demonstrate substantially improved ranked retrieval performance with super-fine attributes in comparison to conventional binary labels, reporting up to a 11.2 and 14.8 percent mAP improvement for gender and age, further surpassed by ethnicity. We also find our 3 super-fine traits to outperform 35 binary attributes by 6.5 percent mAP for subject retrieval in a challenging zero-shot identification scenario.
IEEE transactions on visualization and computer graphics [IEEE Trans Vis Comput Graph] 2018 Dec; Vol. 24 (12), pp. 3044-3057. Date of Electronic Publication: 2017 Dec 18.
We present a novel approach to using Bounding Volume Hierarchies (BVHs) for collision detection of volumetric meshes for digital prototyping based on accurate simulation. In general, volumetric meshes contain more primitives than surface meshes, which in turn means larger BVHs. To manage these larger BVHs, we propose an algorithm for splitting meshes into smaller chunks with a limited-size BVH each. Limited-height BVHs make guided, all-pairs testing of two chunked meshes well-suited for GPU implementation. This is because the dynamically generated work during BVH traversal becomes bounded. Chunking is simple to implement compared to dynamic load balancing methods and can result in an overall two orders of magnitude speedup on GPUs. This indicates that dynamic load balancing may not be a well suited scheme for the GPU. The overall application timings showed that data transfers were not the bottleneck. Instead, the conversion to and from OpenCL friendly data structures was causing serious performance impediments. Still, a simple OpenMP acceleration of the conversion allowed the GPU solution to beat the CPU solution by 20 percent. We demonstrate our results using rigid and deformable body scenes of varying complexities on a variety of GPUs.
IEEE transactions on visualization and computer graphics [IEEE Trans Vis Comput Graph] 2015 Feb; Vol. 21 (2), pp. 201-14.
Visualization applications nowadays not only face increasingly larger datasets, but have to solve increasingly complex research questions. They often require more than a single algorithm and consequently a software solution will exceed the possibilities of simple research prototypes. Well-established systems intended for such complex visual analysis purposes have usually been designed for classical, mesh-based graphics approaches. For particle-based data, however, existing visualization frameworks are too generic - e.g. lacking possibilities for consistent low-level GPU optimization for high-performance graphics - and at the same time are too limited - e.g. by enforcing the use of structures suboptimal for some computations. Thus, we developed the system softwareMegaMol for visualization research on particle-based data. On the one hand, flexible data structures and functional module design allow for easy adaption to changing research questions, e.g. studying vapors in thermodynamics, solid material in physics, or complex functional macromolecules like proteins in biochemistry. Therefore, MegaMol is designed as a development framework. On the other hand, common functionality for data handling and advanced rendering implementations are available and beneficial for all applications. We present several case studies of work implemented using our system as well as a comparison to other freely available or open source systems.
IEEE transactions on visualization and computer graphics [IEEE Trans Vis Comput Graph] 2015 Feb; Vol. 21 (2), pp. 188-200.
Realistic visualization of cloth has many applications in computer graphics. An ongoing research problem is how to best represent and capture cloth models, specifically when considering computer aided design of cloth. Previous methods produce highly realistic images, however, they are either difficult to edit or require the measurement of large databases to capture all variations of a cloth sample. We propose a pipeline to reverse engineer cloth and estimate a parametrized cloth model from a single image. We introduce a geometric yarn model, integrating state-of-the-art textile research. We present an automatic analysis approach to estimate yarn paths, yarn widths, their variation and a weave pattern. Several examples demonstrate that we are able to model the appearance of the original cloth sample. Properties derived from the input image give a physically plausible basis that is fully editable using a few intuitive parameters.
IEEE transactions on visualization and computer graphics [IEEE Trans Vis Comput Graph] 2014 Jul; Vol. 20 (7), pp. 1035-47.
Computer Simulation, Humans, Image Enhancement methods, Image Interpretation, Computer-Assisted methods, Pilot Projects, Reproducibility of Results, Sensitivity and Specificity, Algorithms, Computer Graphics, Imaging, Three-Dimensional methods, Models, Biological, Software, and Whole Body Imaging methods
We present ADAPT, a flexible platform for designing and authoring functional, purposeful human characters in a rich virtual environment. Our framework incorporates character animation, navigation, and behavior with modular interchangeable components to produce narrative scenes. The animation system provides locomotion, reaching, gaze tracking, gesturing, sitting, and reactions to external physical forces, and can easily be extended with more functionality due to a decoupled, modular structure. The navigation component allows characters to maneuver through a complex environment with predictive steering for dynamic obstacle avoidance. Finally, our behavior framework allows a user to fully leverage a character's animation and navigation capabilities when authoring both individual decision-making and complex interactions between actors using a centralized, event-driven model.