arXiv subjects

James Charles

Publications and source records attributed to James Charles.

17 recordsLinked to original sources

arXiv2023

FOUND: Foot Optimization with Uncertain Normals for Surface Deformation Using Synthetic Data

Surface reconstruction from multi-view images is a challenging task, with solutions often requiring a large number of sampled images with high overlap. We seek to develop a method for few-view reconstruction, for the case of the human foot. To solve this task, we must extract rich geometric cues from RGB images, before carefully fusing them into a final 3D object. Our FOUND approach tackles this, with 4 main contributions: (i) SynFoot, a synthetic dataset of 50,000 photorealistic foot images, paired with ground truth surface normals and keypoints; (ii) an uncertainty-aware surface normal predictor trained on our synthetic dataset; (iii) an optimization scheme for fitting a generative foot model to a series of images; and (iv) a benchmark dataset of calibrated images and high resolution ground truth geometry. We show that our normal predictor outperforms all off-the-shelf equivalents significantly on real images, and our optimization scheme outperforms state-of-the-art photogrammetry pipelines, especially for a few-view setting. We release our synthetic dataset and baseline 3D scans to the research community.

cs.CV↗

arXiv2022

FIND: An Unsupervised Implicit 3D Model of Articulated Human Feet

In this paper we present a high fidelity and articulated 3D human foot model. The model is parameterised by a disentangled latent code in terms of shape, texture and articulated pose. While high fidelity models are typically created with strong supervision such as 3D keypoint correspondences or pre-registration, we focus on the difficult case of little to no annotation. To this end, we make the following contributions: (i) we develop a Foot Implicit Neural Deformation field model, named FIND, capable of tailoring explicit meshes at any resolution i.e. for low or high powered devices; (ii) an approach for training our model in various modes of weak supervision with progressively better disentanglement as more labels, such as pose categories, are provided; (iii) a novel unsupervised part-based loss for fitting our model to 2D images which is better than traditional photometric or silhouette losses; (iv) finally, we release a new dataset of high resolution 3D human foot scans, Foot3D. On this dataset, we show our model outperforms a strong PCA implementation trained on the same data in terms of shape quality and part correspondences, and that our novel unsupervised part-based loss improves inference on images.

cs.CV↗

arXiv2021

Discrete neural representations for explainable anomaly detection

The aim of this work is to detect and automatically generate high-level explanations of anomalous events in video. Understanding the cause of an anomalous event is crucial as the required response is dependant on its nature and severity. Recent works typically use object or action classifier to detect and provide labels for anomalous events. However, this constrains detection systems to a finite set of known classes and prevents generalisation to unknown objects or behaviours. Here we show how to robustly detect anomalies without the use of object or action classifiers yet still recover the high level reason behind the event. We make the following contributions: (1) a method using saliency maps to decouple the explanation of anomalous events from object and action classifiers, (2) show how to improve the quality of saliency maps using a novel neural architecture for learning discrete representations of video by predicting future frames and (3) beat the state-of-the-art anomaly explanation methods by 60\% on a subset of the public benchmark X-MAN dataset.

cs.CV↗

arXiv2021

Band tail formation in mono and multilayered transition metal dichalcogenides: A detailed assessment and a quick-reference guide

Transition metal dichalcogenides (TMDs) are promising candidates for a wide variety of ultrascaled electronic, quantum computation, and optoelectronic applications. The exponential decay of electronic density of states into the bandgap, i.e. the band tail has a strong impact on the performance of TMD applications. In this work, the band tails of various TMD monolayer and multilayer systems when placed on various dielectric substrates is predicted with density functional theory based nonequilibrium Green's functions. Nonlocal scattering of electrons on polar optical phonons, charged impurities and remote scattering on phonons in the dielectric materials is included in the self-consistent Born approximation. The band tails are found to critically depend on the layer thickness, temperature, doping concentration and particularly on the chosen dielectric substrate. The underlying physical mechanisms are studied in high detail and an analytical interpolation formula is given to provide a quick-reference for Urbach parameters in $MoS_2$, $WS_2$ and $WSe_2$.

cond-mat.mes-hall↗

arXiv2021

X-MAN: Explaining multiple sources of anomalies in video

Our objective is to detect anomalies in video while also automatically explaining the reason behind the detector's response. In a practical sense, explainability is crucial for this task as the required response to an anomaly depends on its nature and severity. However, most leading methods (based on deep neural networks) are not interpretable and hide the decision making process in uninterpretable feature representations. In an effort to tackle this problem we make the following contributions: (1) we show how to build interpretable feature representations suitable for detecting anomalies with state of the art performance, (2) we propose an interpretable probabilistic anomaly detector which can describe the reason behind it's response using high level concepts, (3) we are the first to directly consider object interactions for anomaly detection and (4) we propose a new task of explaining anomalies and release a large dataset for evaluating methods on this task. Our method competes well with the state of the art on public datasets while also providing anomaly explanation based on objects and their interactions.

cs.CV↗

arXiv2020

Who Left the Dogs Out? 3D Animal Reconstruction with Expectation Maximization in the Loop

We introduce an automatic, end-to-end method for recovering the 3D pose and shape of dogs from monocular internet images. The large variation in shape between dog breeds, significant occlusion and low quality of internet images makes this a challenging problem. We learn a richer prior over shapes than previous work, which helps regularize parameter estimation. We demonstrate results on the Stanford Dog dataset, an 'in the wild' dataset of 20,580 dog images for which we have collected 2D joint and silhouette annotations to split for training and evaluation. In order to capture the large shape variety of dogs, we show that the natural variation in the 2D dataset is enough to learn a detailed 3D prior through expectation maximization (EM). As a by-product of training, we generate a new parameterized model (including limb scaling) SMBLD which we release alongside our new annotation dataset StanfordExtra to the research community.

cs.CV↗

arXiv2020

Mode-space-compatible inelastic scattering in atomistic nonequilibrium Green's function implementations

The nonequilibrium Green's function (NEGF) method is often used to predict transport in atomistically resolved nanodevices and yields an immense numerical load when inelastic scattering on phonons is included. To ease this load, this work extends the atomistic mode space approach of Ref. [1] to include inelastic scattering on optical and acoustic phonons in silicon nanowires. This work also includes the exact calculation of the real part of retarded scattering self-energies in the reduced basis representation using the Kramers-Kronig relations. The inclusion of the Kramers-Kronig relation for the real part of the retarded scattering self-energy increases the impact of scattering. Virtually perfect agreement with results of the original representation is achieved with matrix rank reductions of more than 97%. Time-to-solution improvements of more than 200$\times$ and peak memory reductions of more than 7$\times$ are shown. This allows for the solution of electron transport scattered on phonons in atomically resolved nanowires with cross-sections larger than 5 nm $\times$ 5 nm.

physics.comp-ph↗

arXiv2020

Introducing Open boundary conditions in modeling nonperiodic materials and interfaces: the impact of the periodic assumption

Simulations are essential to accelerate the discovery of new materials and to gain full understanding of known ones. Although hard to realize experimentally, periodic boundary conditions are omnipresent in material simulations. In this work, we intro-duce ROBIN (recursive open boundary and interfaces), the first method allowing open boundary conditions in material and interface modeling. The computational costs are limited to solving quantum properties in a focus area which allows explicitly discretizing millions of atoms in real space and to consider virtually any type of environment (be it periodic, regular, or ran-dom). The impact of the periodicity assumption is assessed in detail with silicon dopants in graphene. Graphene was con-firmed to produce a band gap with periodic substitution of 3% carbon with silicon in agreement with published periodic boundary condition calculations. Instead, 3% randomly distributed silicon in graphene only shifts the energy spectrum. The predicted shift agrees quantitatively with published experimental data. Key insight of this assessment is, assuming periodici-ty elevates a small perturbation of a periodic cell into a strong impact on the material property prediction. Periodic boundary conditions can be applied on truly periodic systems only. More general systems should apply an open boundary method for reliable predictions.

cond-mat.mtrl-sci↗

arXiv2019

Non-equilibrium Green's function predictions of band tails and band gap narrowing in III-V semiconductors and nanodevices

High-doping induced Urbach tails and band gap narrowing play a significant role in determining the performance of tunneling devices and optoelectronic devices such as tunnel field-effect transistors (TFETs), Esaki diodes and light-emitting diodes. In this work, Urbach tails and band gap narrowing values are calculated explicitly for GaAs, InAs, GaSb and GaN as well as ultra-thin bodies and nanowires of the same. Electrons are solved in the non-equilibrium Green's function method in multi-band atomistic tight binding. Scattering on polar optical phonons and charged impurities is solved in the self-consistent Born approximation. The corresponding nonlocal scattering self-energies as well as their numerically efficient formulations are introduced for ultra-thin bodies and nanowires. Predicted Urbach band tails and conduction band gap narrowing agree well with experimental literature for a range of temperatures and doping concentrations. Polynomial fits of the Urbach tail and band gap narrowing as a function of doping are tabulated for quick reference.

cond-mat.mes-hall↗

arXiv2018

Explicit screening full band quantum transport model for semiconductor nanodevices

State of the art quantum transport models for semiconductor nanodevices attribute negative (positive) unit charges to states of the conduction (valence) band. Hybrid states that enable band-to-band tunneling are subject to interpolation that yield model dependent charge contributions. In any nanodevice structure, these models rely on device and physics specific input for the dielectric constants. This paper exemplifies the large variability of different charge interpretation models when applied to ultrathin body transistor performance predictions. To solve this modeling challenge, an electron-only band structure model is extended to atomistic quantum transport. Performance predictions of MOSFETs and tunneling FETs confirm the generality of the new model and its independence of additional screening models.

physics.app-ph↗

arXiv2017

Control of interlayer delocalization in 2H transition metal dichalcogenides

It is assessed in detail both experimentally and theoretically how the interlayer coupling of transition metal dichalcogenides controls the electronic properties of the respective devices. Gated transition metal dichalcogenide structures show electrons and holes to either localize in individual monolayers, or delocalize beyond multiple layers - depending on the balance between spin-orbit inter- action and interlayer hopping. This balance depends on layer thickness, momentum space symmetry points and applied gate fields. A good quantitative agreement of predictions and measurements of the quantum confined Stark effect in gated MoS2 systems unveils intralayer excitons as major source for the observed photoluminesence.

cond-mat.mes-hall↗

arXiv2017

Thermal Transport Across Metal Silicide-Silicon Interfaces: First-Principles Calculations and Green's Function Transport Simulations

In this work, we use a combination of first-principles calculations under the density functional theory framework and heat transport simulations using the atomistic Green's function (AGF) method to quantitatively predict the contribution of the different scattering mechanisms to the thermal interface conductance of epitaxial CoSi$_2$-Si interfaces. An important development in the present work is the direct computation of interfacial bonding from density functional perturbation theory (DFPT) and hence the avoidance of commonly used `mixing rules' to obtain the cross-interface force constants from bulk material force constants. Another important algorithmic development is the integration of the recursive Green's function (RGF) method with Büttiker probe scattering that enables computationally efficient simulations of inelastic phonon scattering and its contribution to the thermal interface conductance. First-principles calculations of electron-phonon coupling reveal that cross-interface energy transfer between metal electrons and atomic vibrations in the semiconductor is mediated by delocalized acoustic phonon modes that extend on both sides of the interface, and phonon modes that are localized inside the semiconductor region of the interface exhibit negligible coupling with electrons in the metal. We also provide a direct comparison between simulation predictions and experimental measurements of thermal interface conductance of epitaxial CoSi$_2$-Si interfaces using the time-domain thermoreflectance technique. Importantly, the experimental results, performed across a wide temperature range, only agree well with predictions that include all transport processes: elastic and inelastic phonon scattering, electron-phonon coupling in the metal, and electron-phonon coupling across the interface.

cond-mat.mes-hall↗

arXiv2016

Personalizing Human Video Pose Estimation

We propose a personalized ConvNet pose estimator that automatically adapts itself to the uniqueness of a person's appearance to improve pose estimation in long videos. We make the following contributions: (i) we show that given a few high-precision pose annotations, e.g. from a generic ConvNet pose estimator, additional annotations can be generated throughout the video using a combination of image-based matching for temporally distant frames, and dense optical flow for temporally local frames; (ii) we develop an occlusion aware self-evaluation model that is able to automatically select the high-quality and reject the erroneous additional annotations; and (iii) we demonstrate that these high-quality annotations can be used to fine-tune a ConvNet pose estimator and thereby personalize it to lock on to key discriminative features of the person's appearance. The outcome is a substantial improvement in the pose estimates for the target video using the personalized ConvNet compared to the original generic ConvNet. Our method outperforms the state of the art (including top ConvNet methods) by a large margin on two standard benchmarks, as well as on a new challenging YouTube video dataset. Furthermore, we show that training from the automatically generated annotations can be used to improve the performance of a generic ConvNet on other benchmarks.

cs.CV↗

arXiv2015

Flowing ConvNets for Human Pose Estimation in Videos

The objective of this work is human pose estimation in videos, where multiple frames are available. We investigate a ConvNet architecture that is able to benefit from temporal context by combining information across the multiple frames using optical flow. To this end we propose a network architecture with the following novelties: (i) a deeper network than previously investigated for regressing heatmaps; (ii) spatial fusion layers that learn an implicit spatial model; (iii) optical flow is used to align heatmap predictions from neighbouring frames; and (iv) a final parametric pooling layer which learns to combine the aligned heatmaps into a pooled confidence map. We show that this architecture outperforms a number of others, including one that uses optical flow solely at the input layers, one that regresses joint coordinates directly, and one that predicts heatmaps without spatial fusion. The new architecture outperforms the state of the art by a large margin on three video pose estimation datasets, including the very challenging Poses in the Wild dataset, and outperforms other deep methods that don't use a graphical model on the single-image FLIC benchmark (and also Chen & Yuille and Tompson et al. in the high precision region).

cs.CV↗

arXiv2015

NEMO5: Achieving High-end Internode Communication for Performance Projection Beyond Moore's Law

Electronic performance predictions of modern nanotransistors require nonequilibrium Green's functions including incoherent scattering on phonons as well as inclusion of random alloy disorder and surface roughness effects. The solution of all these effects is numerically extremely expensive and has to be done on the world's largest supercomputers due to the large memory requirement and the high performance demands on the communication network between the compute nodes. In this work, it is shown that NEMO5 covers all required physical effects and their combination. Furthermore, it is also shown that NEMO5's implementation of the algorithm scales very well up to about 178176CPUs with a sustained performance of about 857 TFLOPS. Therefore, NEMO5 is ready to simulate future nanotransistors.

cs.DC↗

arXiv2015

Theoretical study of strain-dependent optical absorption in Stranski-Krastanov grown InAs/InGaAs/GaAs/AlGaAs quantum dots

A detailed theoretical study of the optical absorption in self-assembled quantum dots is presented in this paper. A rigorous atomistic strain model as well as a sophisticated electronic band structure model are used to ensure accurate prediction of the optical transitions in these devices . The optimized models presented in this paper are able to reproduce the experimental results with an error less than 1$\%$. The effects of incident light polarization, alloy mole fraction, quantum dot dimensions, and doping have been investigated. The in-plane polarized light absorption is more significant than the perpendicularly polarized light absorption. Increasing the mole fraction of the strain controlling layer leads to a lower energy gap and larger absorption wavelength. Surprisingly, the absorption wavelength is highly sensitive to changes in the dot diameter, but almost insensitive to changes in the dot height. This unpredicted behavior is explained by sensitivity analysis of different factors which affect the optical transition energy.

cond-mat.mes-hall↗

arXiv2013

An Environment-dependent Semi-Empirical Tight Binding Model Suitable for Electron Transport in Bulk Metals, Metal Alloys, Metallic Interfaces and Metallic Nanostructures II - Effect of Confinement and Homogeneous Strain on Cu Conductance

The Semi-Empirical TB model developed in part I is applied to metal transport problems of current relevance in part II. A systematic study of the effect of quantum confinement, transport orientation and homogeneous strain on electronic transport properties of Cu is carried out. It is found that quantum confinement from bulk to nanowire boundary conditions leads to significant anisotropy in conductance of Cu along different transport orientations. Compressive homogeneous strain is found to reduce resistivity by increasing the density of conducting modes in Cu. The [110] transport orientation in Cu nanowires is found to be the most favorable for mitigating conductivity degradation since it shows least reduction in conductance with confinement and responds most favorably to compressive strain.

cond-mat.mtrl-sci↗