60419 research outputs found
Sort by
Semantic Estimation of 3D Body Shape and Pose using Minimal Cameras
No full textWe aim to simultaneously estimate the 3D articulated pose and high fidelity volumetric
occupancy of human performance, from multiple viewpoint video (MVV) with as
few as two views. We use a multi-channel symmetric 3D convolutional encoder-decoder
with a dual loss to enforce the learning of a latent embedding that enables inference of
skeletal joint positions and a volumetric reconstruction of the performance. The inference
is regularised via a prior learned over a dataset of view-ablated multi-view video footage
of a wide range of subjects and actions, and show this to generalise well across unseen
subjects and actions. We demonstrate improved reconstruction accuracy and lower pose
estimation error relative to prior work on two MVV performance capture datasets: Human
3.6M and TotalCapture
Modeling Label Dependencies for Audio Tagging with Graph Convolutional Network
No full textAs a multi-label classification task, audio tagging
aims to predict the presence or absence of certain sound events in an audio recording. Existing works in audio tagging do not explicitly consider the probabilities of the co-occurrences between
sound events, which is termed as the label dependencies in this study. To address this issue, we propose to model the label dependencies via a graph-based method, where each node of the graph represents a label. An adjacency matrix is constructed
by mining the statistical relations between labels to represent the
graph structure information, and a graph convolutional network
(GCN) is employed to learn node representations by propagating
information between neighboring nodes based on the adjacency
matrix, which implicitly models the label dependencies. The
generated node representations are then applied to the acoustic
representations for classification. Experiments on Audioset
show that our method achieves a state-of-the-art mean average
precision (mAP) of 0:434
Detailed Abundances in the Ultra-faint Magellanic Satellites Carina II and III
No full textWe present the first detailed elemental abundances in the ultra-faint Magellanic satellite galaxies Carina II (Car II) and Carina III (Car III). With high-resolution Magellan/MIKE spectroscopy, we determined the abundances of nine stars in Car II, including the first abundances of an RR Lyrae star in an ultra-faint dwarf galaxy (UFD), and two stars in Car III. The chemical abundances demonstrate that both systems are clearly galaxies and not globular clusters. The stars in these galaxies mostly display abundance trends matching those of other similarly faint dwarf galaxies: enhanced but declining [α/Fe] ratios, iron-peak elements matching the stellar halo, and unusually low neutron-capture element abundances. One star displays a low outlying [Sc/Fe] = −1.0. We detect a large Ba scatter in Car II, likely due to inhomogeneous enrichment by low-mass asymptotic giant branch star winds. The most striking abundance trend is for [Mg/Ca] in Car II, which decreases from +0.4 to −0.4 and indicates clear variation in the initial progenitor masses of enriching core-collapse supernovae. So far, the only UFDs displaying a similar [Mg/Ca] trend are likely satellites of the Large Magellanic Cloud. We find two stars with [Fe/H] ≤ −3.5 whose abundances likely trace the first generation of metal-free Population III stars and are well fit by Population III core-collapse supernova yields. An appendix describes our new abundance uncertainty analysis that propagates line-by-line stellar parameter uncertainties
Sketchformer: Transformer-Based Representation for Sketched Structure
No full textSketchformer is a novel transformer-based representation
for encoding free-hand sketches input in a vector form,
i.e. as a sequence of strokes. Sketchformer effectively addresses
multiple tasks: sketch classification, sketch based
image retrieval (SBIR), and the reconstruction and interpolation
of sketches. We report several variants exploring continuous
and tokenized input representations, and contrast
their performance. Our learned embedding, driven by a
dictionary learning tokenization scheme, yields state of the
art performance in classification and image retrieval tasks,
when compared against baseline representations driven by
LSTM sequence to sequence architectures: SketchRNN and
derivatives. We show that sketch reconstruction and interpolation
are improved significantly by the Sketchformer embedding
for complex sketches with longer stroke sequences
What are we measuring with the Morningness-Eveningness Questionnaire? Exploratory factor analysis across four samples from two countries
No full textIndividual variability in diurnal preference or chronotype is commonly assessed with selfreport
scales such as the widely used Morningness-Eveningness Questionnaire (MEQ).
We sought to investigate the MEQ’s internal consistency by applying exploratory factor
analysis (EFA) to determine the number of underlying latent factors in four different adult
samples, two each from the United Kingdom and Brazil (total N=3,457). We focused on
factors that were apparent in all samples, irrespective of particular sociocultural diversity
and geographical characteristics, so as to show a common core reproducible structure
across samples. Results showed a three-factor solution with acceptable to good model fit
indexes in all studied populations. Twelve of the 19 MEQ items in the three-correlated
factor solution loaded onto the same factors across the four samples. This shows that the
scale measures three distinguishable, yet correlated constructs: 1) items related to how
people feel in the morning, which we termed efficiency of dissipation of sleep pressure
(recovery process) (items 1, 3, 4, 5, 7, 9, 13, and 19); 2) items related to how people feel
before sleep, which we called sensitivity to build-up of sleep pressure (items 2, 10, and
12); and 3) peak time of cognitive arousal (item 11). Although the third factor was not
regarded as consistent since only one item was common among all samples, it might
represent subjective amplitude. These results suggested that the latent constructs of the
MEQ reflect dissociable homeostatic processes in addition to a less consistent propensity
for cognitive arousal at different times of the day. By analysing answers to MEQ items
that compose these latent factors, it may be possible to extract further knowledge of
factors that affect morningness-eveningness
Statistical limitations in proton imaging
No full textProton imaging is a promising technology for proton radiotherapy as it can be used for: (1) direct sampling of the tissue stopping power, (2) input information for multi-modality RSP reconstruction, (3) gold-standard calibration against concurrent techniques, (4) tracking motion and (5) pre-treatment positioning. However, no end-to-end characterization of the image quality (signal-to-noise ratio and spatial resolution, blurring uncertainty) against the dose has been done. This work aims to establish a model relating these characteristics and to describe their relationship with proton energy and object size. The imaging noise originates from two processes: the Coulomb scattering with the nucleus, producing a path deviation, and the energy loss straggling with electrons. The noise is found to increases with thickness crossed and, independently, decreases with decreasing energy. The scattering noise is dominant around high-gradient edge whereas the straggling noise is maximal in homogeneous regions. Image quality metrics are found to behave oppositely against energy: lower energy minimizes both the noise and the spatial resolution, with the optimal energy choice depending on the application and location in the imaged object. In conclusion, the model presented will help define an optimal usage of proton imaging to reach the promised application of this technology and establish a fair comparison with other imaging techniques
Front-of-pack images can boost the perceived health benefits of dietary products
No full textImages on dietary supplement packaging can help identify the products' supposed function. However, research shows that these images can also lead people to infer additional health benefits of consuming the products. The present research investigated the extent to which front-of-pack imagery affects people's perceptions of the health risks and benefits of fictional products. In three randomized experiments, participants saw fictitious dietary supplement packages. Some of the packages included a health-related image (e.g. a heart), whereas others did not. Participants were asked to infer the products' intended purpose and then to rate the perceived risks and benefits of consuming the product. In Experiment 1 (N = 546), the inclusion of a health-related image increased the perceived benefits of consuming the product, with minimal effect on the perceived risks. This finding was replicated in Experiment 2 (N = 164), but was contingent on whether each product's assumed health function was confirmed or disconfirmed. In Experiment 3 (N = 306), which used a pre-registered design and analysis plan, the inclusion of a health-related image increased the perceived benefits and decreased the perceived risks of consuming the product. Again, these effects were contingent on whether the assumed health functions were confirmed or disconfirmed. These findings indicate that health-related imagery could lead consumers to infer additional health properties from non-diagnostic information featured on a product's packaging, perhaps as a consequence of increased processing fluency. This research underscores the importance of regulating the use of imagery in health marketing, to protect consumers from the effects of potentially misleading claims
Temporally coherent general dynamic scene reconstruction
No full textExisting techniques for dynamic scene re-
construction from multiple wide-baseline cameras pri-
marily focus on reconstruction in controlled environ-
ments, with fixed calibrated cameras and strong prior
constraints. This paper introduces a general approach to
obtain a 4D representation of complex dynamic scenes
from multi-view wide-baseline static or moving cam-
eras without prior knowledge of the scene structure, ap-
pearance, or illumination. Contributions of the work
are: An automatic method for initial coarse reconstruc-
tion to initialize joint estimation; Sparse-to-dense tem-
poral correspondence integrated with joint multi-view
segmentation and reconstruction to introduce tempo-
ral coherence; and a general robust approach for joint
segmentation refinement and dense reconstruction of
dynamic scenes by introducing shape constraint. Com-
parison with state-of-the-art approaches on a variety of
complex indoor and outdoor scenes, demonstrates im-
proved accuracy in both multi-view segmentation and
dense reconstruction. This paper demonstrates unsuper-
vised reconstruction of complete temporally coherent
4D scene models with improved non-rigid object seg-
mentation and shape reconstruction and its application
to various applications such as free-view rendering and
virtual reality
Perturbative Quantum Field Theory and Homotopy Algebras
No full textWe review the homotopy algebraic perspective on perturbative quantum field theory: classical field theories correspond to homotopy algebras such as A∞- and L∞-algebras. Furthermore, their scattering amplitudes are encoded in minimal models of these homotopy algebras at tree level and their quantum relatives at loop level. The translation between Lagrangian field theories and homotopy algebras is provided by the Batalin-Vilkovisky formalism. The minimal models are computed recursively using the homological perturbation lemma, which induces useful recursion relations for the computation of scattering amplitudes. After explaining how the homolcogical perturbation lemma produces the usual Feynman diagram expansion, we use our techniques to verify an identity for the Berends-Giele currents which implies the Kleiss-Kuijf relations
Graph Theory assisted Bit-to-Index-Combination Gray Coding for Generalized Index Modulation
No full textGeneralized index modulation (GIM) which implicitly
conveys information by the activated indices is a promising
technique for next-generation wireless networks. Due to the
prohibitive challenge of bit-to-index combination (IC) mapping
optimization, conventional GIM system obtains the bit-to-IC
mapping table randomly, which may suffer from some performance
loss. To circumvent this issue, we propose a low-complexity
graph theory assisted bit-to-IC gray coding for GIM systems by
minimizing the average hamming distance (HD) between any two
ICs having one different value. Specifically, we decompose and
transform the optimization problem into two subproblems using
the graph theory, i.e., 1) Select an IC set whose corresponding
graph has the minimum degree; 2) Design a bit-to-IC mapping
principle to minimize the weight of the selected graph. Low complexity
algorithms are developed to solve the subproblems
with a significant reduced complexity. Both simulation and
theoretical results are shown that the GIM systems with our
proposed mapping table are capable of providing significant
performance gains over the conventional counterparts without
the need for any additional feedback-link and without extra
computational complexity. It is also shown that the proposed
bit-to-IC mapping table is straightforward for any GIM systems
over generalized fading channels
Data sources:
Access Repository Dashboard
Do you manage Surrey Research Insight? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!