arXiv subjects

Wayne Hayes

Publications and source records attributed to Wayne Hayes.

9 recordsLinked to original sources

arXiv2025

PoseBench3D: A Cross-Dataset Analysis Framework for 3D Human Pose Estimation via Pose Lifting Networks

Reliable three-dimensional human pose estimation (3D HPE) remains challenging due to the differences in viewpoints, environments, and camera conventions among datasets. As a result, methods that achieve near-optimal in-dataset accuracy often degrade on unseen datasets. In practice, however, systems must adapt to diverse viewpoints, environments, and camera setups--conditions that differ significantly from those encountered during training, which is often the case in real-world scenarios. Measuring cross-dataset performance is a vital process, but extremely labor-intensive when done manually for human pose estimation. To address these challenges, we automate this evaluation using PoseBench3D, a standardized testing framework that enables consistent and fair cross-dataset comparisons on previously unseen data. PoseBench3D streamlines testing across four widely used 3D HPE datasets via a single, configurable interface. Using this framework, we re-evaluate 18 methods and report over 100 cross-dataset results under Protocol 1: MPJPE and Protocol 2: PA-MPJPE, revealing systematic generalization gaps and the impact of common preprocessing and dataset setup choices. The PoseBench3D code is found at: https://github.com/bryanjvela/PoseBench3D

cs.CV↗

arXiv2023

Learning from Abstract Images: on the Importance of Occlusion in a Minimalist Encoding of Human Poses

Existing 2D-to-3D pose lifting networks suffer from poor performance in cross-dataset benchmarks. Although the use of 2D keypoints joined by "stick-figure" limbs has shown promise as an intermediate step, stick-figures do not account for occlusion information that is often inherent in an image. In this paper, we propose a novel representation using opaque 3D limbs that preserves occlusion information while implicitly encoding joint locations. Crucially, when training on data with accurate three-dimensional keypoints and without part-maps, this representation allows training on abstract synthetic images, with occlusion, from as many synthetic viewpoints as desired. The result is a pose defined by limb angles rather than joint positions $\unicode{x2013}$ because poses are, in the real world, independent of cameras $\unicode{x2013}$ allowing us to predict poses that are completely independent of camera viewpoint. The result provides not only an improvement in same-dataset benchmarks, but a "quantum leap" in cross-dataset benchmarks.

cs.CV↗

arXiv2022

Detection of large exact subgraph isomorphisms with a topology-only graphlet index built using deterministic walks

We introduce the first algorithm to perform topology-only local graph matching (a.k.a. local network alignment or subgraph isomorphism): BLANT, for Basic Local Alignment of Network Topology. BLANT first creates a limited, high-specificity index of a single graph containing connected k-node induced subgraphs called k-graphlets, for k=6-15. The index is constructed in a deterministic way such that, if significant common network topology exists between two networks, their indexes are likely to overlap. This is the key insight which allows BLANT to discover alignments using only topological information. To align two networks, BLANT queries their respective indexes to form large, high quality local alignments. BLANT is able to discover highly topologically similar alignments (S3 >= 0.95) of up to 150 node-pairs for which up to 50% of node pairs differ from their "assigned" global counterpart. These results compare favorably against the baseline, a state-of-the-art local alignment algorithm which was adapted to be topology-only. Such alignments are 3x larger and differ 30% more (additive) more from the global alignment than alignments of similar topological similarity (S3 >= 0.95) discovered by the baseline. We hope that such regions of high local similarity and low global similarity may provide complementary insights to global alignment algorithms.

cs.SI↗

arXiv2016

On the nature and correction of the spurious S-wise spiral galaxy winding bias in Galaxy Zoo 1

The Galaxy Zoo 1 catalog displays a bias towards the S-wise winding direction in spiral galaxies which has yet to be explained. The lack of an explanation confounds our attempts to verify the Cosmological Principle, and has spurred some debate as to whether a bias exists in the real universe. The bias manifests not only in the obvious case of trying to decide if the universe as a whole has a winding bias, but also in the more insidious case of selecting which galaxies to include in a winding direction survey. While the former bias has been accounted for in a previous image-mirroring study, the latter has not. Furthermore, the bias has never been {\em corrected} in the GZ1 catalog, as only a small sample of the GZ1 catalog was re-examined during the mirror study. We show that the existing bias is a human {\em selection} effect rather than a human chirality bias. In effect, the excess S-wise votes are spuriously "stolen" from the elliptical and edge-on-disk categories, not the Z-wise category. Thus, when selecting a set of spiral galaxies by imposing a threshold $T$ so that $\max(P_S,P_Z) > T$ or $P_S+P_Z>T$, we spuriously select more S-wise than Z-wise galaxies. We show that when a provably unbiased machine selects which galaxies are spirals independent of their chirality, the S-wise surplus vanishes, even if humans are still used to determine the chirality. Thus, when viewed across the entire GZ1 sample (and by implication, the Sloan catalog), the winding direction of arms in spiral galaxies as viewed from Earth is consistent with the flip of a fair coin.

astro-ph.GA↗

arXiv2016

SANA: Simulated Annealing Network Alignment Applied to Biological Networks

The alignment of biological networks has the potential to teach us as much about biology and disease as has sequence alignment. Sequence alignment can be optimally solved in polynomial time. In contrast, network alignment is $NP$-hard, meaning optimal solutions are impossible to find, and the quality of found alignments depend strongly upon the algorithm used to create them. Every network alignment algorithm consists of two orthogonal components: first, an objective function or measure $M$ that is used to evaluate the quality of any proposed alignment, and second, a search algorithm used to explore the exponentially large set of possible alignments in an effort to find "good" ones according to $M$. Objective functions fall into many categories, including biological measures such as sequence similarity, as well as topological measures like graphlet similarity and edge coverage (possibly weighted). Algorithms to search the space of all possible alignments can be deterministic or stochastic, and many possibilities have been offered over the past decade. In this paper we introduce a new stochastic search algorithm called SANA: Simulated Annealing Network Aligner. We test it on several popular objective functions and demonstrate that it almost universally optimizes each one significantly better than existing search algorithms. Finally, we compare several topological objective functions using SANA. Software available at http://sana.ics.uci.edu.

q-bio.MN↗

arXiv2013

Automated Quantitative Description of Spiral Galaxy Arm-Segment Structure

We describe a system for the automatic quantification of structure in spiral galaxies. This enables translation of sky survey images into data needed to help address fundamental astrophysical questions such as the origin of spiral structure---a phenomenon that has eluded theoretical description despite 150 years of study (Sellwood 2010). The difficulty of automated measurement is underscored by the fact that, to date, only manual efforts (such as the citizen science project Galaxy Zoo) have been able to extract information about large samples of spiral galaxies. An automated approach will be needed to eliminate measurement subjectivity and handle the otherwise-overwhelming image quantities (up to billions of images) from near-future surveys. Our approach automatically describes spiral galaxy structure as a set of arcs, precisely describing spiral arm segment arrangement while retaining the flexibility needed to accommodate the observed wide variety of spiral galaxy structure. The largest existing quantitative measurements were manually-guided and encompassed fewer than 100 galaxies, while we have already applied our method to more than 29,000 galaxies. Our output matches previous information, both quantitatively over small existing samples, and qualitatively against human classifications from Galaxy Zoo.

astro-ph.CO↗

arXiv2009

Topological network alignment uncovers biological function and phylogeny

Sequence comparison and alignment has had an enormous impact on our understanding of evolution, biology, and disease. Comparison and alignment of biological networks will likely have a similar impact. Existing network alignments use information external to the networks, such as sequence, because no good algorithm for purely topological alignment has yet been devised. In this paper, we present a novel algorithm based solely on network topology, that can be used to align any two networks. We apply it to biological networks to produce by far the most complete topological alignments of biological networks to date. We demonstrate that both species phylogeny and detailed biological function of individual proteins can be extracted from our alignments. Topology-based alignments have the potential to provide a completely new, independent source of phylogenetic information. Our alignment of the protein-protein interaction networks of two very different species--yeast and human--indicate that even distant species share a surprising amount of network topology with each other, suggesting broad similarities in internal cellular wiring across all life on Earth.

q-bio.MN↗

arXiv2002

Shadowing-based reliability decay in softened n-body simulations

A shadow of a numerical solution to a chaotic system is an_exact_ solution to the equations of motion that remains close to the numerical solution for a long time. In a collisionless n-body system, we know that particle motion is governed by the global potential rather than by inter-particle interactions. As a result, the trajectory of each individual particle in the system is independently shadowable. It is thus meaningful to measure the number of particles that have shadowable trajectories as a function of time. We find that the number of shadowable particles decays exponentially with time as exp(-mu t), and that for eps in [~0.2,1] (in units of the local mean inter-particle separation $\bar n$), there is an explicit relationship between the decay constant mu, the timestep h of the leapfrog integrator, the softening eps, and the number of particles N in the simulation. Thus, given N and eps, it is possible to pre-compute the timestep h necessary to acheive a desired fraction of shadowable particles after a given length of simulation time. We demonstrate that a large fraction of particles remain shadowable over ~100 crossing times even if particles travel up to about 1/3 of the softening length per timestep. However, a sharp decrease in the number of shadowable particles occurs if the timestep increases to allow particles to travel further than 1/3 the softening length in one timestep, or if the softening is decreased below ~0.2$\bar n$.

astro-ph↗

arXiv1997

Fitting random stable solar systems to Titius-Bode laws

Simple ``solar systems'' are generated with planetary orbital radii r distributed uniformly random in log(r) between 0.2 and 50 AU. A conservative stability criterion is imposed by requiring that adjacent planets are separated by a minimum distance of k Hill radii, for values of k ranging from 1 to 8. Least-squares fits of these systems to generalized Bode laws are performed, and compared to the fit of our own Solar System. We find that this stability criterion, and other ``radius-exclusion'' laws, generally produce approximately geometrically spaced planets that fit a Titius-Bode law about as well as our own Solar System. We then allow the random systems the same exceptions that have historically been applied to our own Solar System. Namely, one gap may be inserted, similar to the gap between Mars and Jupiter, and up to 3 planets may be ``ignored'', similar to how some forms of Bode's law ignore Mercury, Neptune, and Pluto. With these particular exceptions, we find that our Solar System fits significantly better than the random ones. However, we believe that this choice of exceptions, designed specifically to give our own Solar System a better fit, gives it an unfair advantage that would be lost if other exception rules were used. We conclude that the significance of Bode's law is simply that stable planetary systems tend to be regularly spaced; this conclusion could be strengthened by the use of more stringent methods of rejecting unstable solar systems, such as long-term orbit integrations.

astro-ph↗