arXiv subjects

Christopher Thomas

Publications and source records attributed to Christopher Thomas.

32 records · Page 2Linked to original sources

arXiv2018

Artistic Object Recognition by Unsupervised Style Adaptation

Computer vision systems currently lack the ability to reliably recognize artistically rendered objects, especially when such data is limited. In this paper, we propose a method for recognizing objects in artistic modalities (such as paintings, cartoons, or sketches), without requiring any labeled data from those modalities. Our method explicitly accounts for stylistic domain shifts between and within domains. To do so, we introduce a complementary training modality constructed to be similar in artistic style to the target domain, and enforce that the network learns features that are invariant between the two training modalities. We show how such artificial labeled source domains can be generated automatically through the use of style transfer techniques, using diverse target images to represent the style in the target domain. Unlike existing methods which require a large amount of unlabeled target data, our method can work with as few as ten unlabeled images. We evaluate it on a number of cross-domain object and scene classification tasks and on a new dataset we release. Our experiments show that our approach, though conceptually simple, significantly improves the accuracy that existing domain adaptation techniques obtain for artistic object recognition.

cs.CV↗

arXiv2018

Persuasive Faces: Generating Faces in Advertisements

In this paper, we examine the visual variability of objects across different ad categories, i.e. what causes an advertisement to be visually persuasive. We focus on modeling and generating faces which appear to come from different types of ads. For example, if faces in beauty ads tend to be women wearing lipstick, a generative model should portray this distinct visual appearance. Training generative models which capture such category-specific differences is challenging because of the highly diverse appearance of faces in ads and the relatively limited amount of available training data. To address these problems, we propose a conditional variational autoencoder which makes use of predicted semantic attributes and facial expressions as a supervisory signal when training. We show how our model can be used to produce visually distinct faces which appear to be from a fixed ad topic category. Our human studies and quantitative and qualitative experiments confirm that our method greatly outperforms a variety of baselines, including two variations of a state-of-the-art generative adversarial network, for transforming faces to be more ad-category appropriate. Finally, we show preliminary generation results for other types of objects, conditioned on an ad topic.

cs.CV↗

arXiv2017

Automatic Understanding of Image and Video Advertisements

There is more to images than their objective physical content: for example, advertisements are created to persuade a viewer to take a certain action. We propose the novel problem of automatic advertisement understanding. To enable research on this problem, we create two datasets: an image dataset of 64,832 image ads, and a video dataset of 3,477 ads. Our data contains rich annotations encompassing the topic and sentiment of the ads, questions and answers describing what actions the viewer is prompted to take and the reasoning that the ad presents to persuade the viewer ("What should I do according to this ad, and why should I do it?"), and symbolic references ads make (e.g. a dove symbolizes peace). We also analyze the most common persuasive strategies ads use, and the capabilities that computer vision systems should have to understand these strategies. We present baseline classification results for several prediction tasks, including automatically answering questions about the messages of the ads.

cs.CV↗

arXiv2016

s- and d-wave superconductivity in a two-band model

Superconductivity in strongly correlated systems is a remarkable phenomenon that attracts a huge interest. The study of this problem is relevant for materials as the high Tc oxides, pnictides and heavy fermions. In this work we study a realistic model that includes the relevant physics of superconductivity in the presence of strong Coulomb correlations. We consider a two-band model, since most of these correlated systems have electrons from at least two different atomic orbitals coexisting at their Fermi surface. The Coulomb repulsion is taken into account through a local repulsive interaction. Pairing is considered among quasi- particles in neighbouring sites and we allow for different symmetries of the order parameter. In order to deal with the strong local correlations, we use the well known slave boson approach that has proved very successful for this problem. Here we are interested in obtaining the zero temperature properties of the model, specifically its phase diagram and the existence and nature of superconducting quantum critical points. We show that these can arise by increasing the mixing between the two bands. Since this can be controlled by external pressure or doping, our results have a direct relation with experiments. We show that the superconductor-to-normal transition can be either to a metal, a correlated metal or to an insulator. Also we compare the relative stability of s and d-wave paired states for different regions of parameter space and investigate the BCS- BEC crossover in the two-band lattice model as function of the strength of the pairing interaction.

cond-mat.supr-con↗

arXiv2016

Seeing Behind the Camera: Identifying the Authorship of a Photograph

We introduce the novel problem of identifying the photographer behind a photograph. To explore the feasibility of current computer vision techniques to address this problem, we created a new dataset of over 180,000 images taken by 41 well-known photographers. Using this dataset, we examined the effectiveness of a variety of features (low and high-level, including CNN features) at identifying the photographer. We also trained a new deep convolutional neural network for this task. Our results show that high-level features greatly outperform low-level features. We provide qualitative results using these learned models that give insight into our method's ability to distinguish between photographers, and allow us to draw interesting conclusions about what specific photographers shoot. We also demonstrate two applications of our method.

cs.CV↗

arXiv2016

OpenSalicon: An Open Source Implementation of the Salicon Saliency Model

In this technical report, we present our publicly downloadable implementation of the SALICON saliency model. At the time of this writing, SALICON is one of the top performing saliency models on the MIT 300 fixation prediction dataset which evaluates how well an algorithm is able to predict where humans would look in a given image. Recently, numerous models have achieved state-of-the-art performance on this benchmark, but none of the top 5 performing models (including SALICON) are available for download. To address this issue, we have created a publicly downloadable implementation of the SALICON model. It is our hope that our model will engender further research in visual attention modeling by providing a baseline for comparison of other algorithms and a platform for extending this implementation. The model we provide supports both training and testing, enabling researchers to quickly fine-tune the model on their own dataset. We also provide a pre-trained model and code for those users who only need to generate saliency maps for images without training their own model.

cs.CV↗

arXiv2015

Hand Posture's Effect on Touch Screen Text Input Behaviors: A Touch Area Based Study

Mobile devices with touch keyboards have become ubiquitous, but text entry on these devices remains slow and errorprone. Understanding touch patterns during text entry could be useful in designing robust error-correction algorithms for soft keyboards. In this paper, we present an analysis of text input behaviors on a soft QWERTY keyboard in three different text entry postures: index finger only, one thumb, and two thumb. Our work expands on the work of [1] by considering the entire surface area of digit contact with the smartphone keyboard, rather than interpreting each touch as a single point. To do this, we captured touch areas for every key in a lab study with 8 participants and calculated offsets, error rates, and size measurements. We then repeated the original experiment described in [1] and showed that significant differences exist when basing offset calculations on touch area compared to touch points for two postures.

cs.HC↗

arXiv2014

Spin liquid versus long range magnetic order in the frustrated body-centered tetragonal lattice

The quantum Heisenberg model is studied in the geometrically frustrated body-centered tetragonal lattice(BCT lattice) with antiferromagnetic interlayer coupling J1 and intralayer first and second neighbor coupling J2 and J3. We introduce a variational method: each interaction term can be decoupled partially in the purely magnetic Weiss and in the spin-liquid (SL) mean-field channels. We find that the most stable variational solutions correspond to the three different possible long range magnetic orders that are respectively governed by J1, J2, and J3. We characterize three different purely SL non-magnetic solutions that are variationally the second most stable states after the purely magnetic ones. This suggests that quantum fluctuations induced by the frustration between J1-J2-J3 coupling should destroy magnetic orders and stabilize the formation of SL in large areas of parameters. The SL solution governed by J1 breaks the lattice translation symmetry. This Modulated SL is associated to a commensurate ordering wave vector (1,1,1). We discuss the relevance of our results for heavy fermion Uru2Si2 and cuprate superconductors that have a BCT lattice structure. Also, the general variational method introduced here can be applied to any other system where interaction terms can be decoupled in two different mean-field channels.

cond-mat.str-el↗

arXiv2014

Effect of anisotropy in the $S=1$ underscreened Kondo lattice

We study the effect of crystal field anisotropy in the underscreened $S=1$ Kondo lattice model. Starting from the two orbital Anderson lattice model and including a local anisotropy term, we show, through Schrieffer-Wolff transformation, that local anisotropy is equivalent to an anisotropic Kondo interaction ($J_{\parallel} \neq{J_{\perp}}$). The competition and coexistence between ferromagnetism and Kondo effect in this effective model is studied within a generalized mean-field approximation. Several regimes are obtained, depending on the parameters, exhibiting or not coexistence of magnetic order and Kondo effect. Particularly, we show that a re-entrant Kondo phase at low temperature can be obtained. We are also able to describe phases where the Kondo temperature is smaller than the Curie temperature ($T_K<T_C$). We propose that some aspects of uranium and neptunium compounds that present coexistence of Kondo effect and ferromagnetism, can be understood within this model.

cond-mat.str-el↗

arXiv2012

3D Modulated Spin Liquid model applied to URu$_2$Si$_2$

We have developed a 3D version for the Modulated Spin Liquid in a body-centered tetragonal lattice structure to describe the hidden order observed in URu$_2$Si$_2$ at $T_0\approx17.5$ K. This second order transition is well described by our model confirming our earlier hypothesis. The symmetry of the modulation is minimized for ${\bf Q}\equiv(1,1,1)$. We assume a linear variation of the interaction parameters with the lattice spacing and our results show good agreement with uniaxial and pressure experiments.

cond-mat.str-el↗

arXiv2010

Application of the S=1 underscreened Anderson lattice model to Kondo uranium and neptunium compounds

Magnetic properties of uranium and neptunium compounds showing the coexistence of Kondo screening effect and ferromagnetic order are investigated within the Anderson lattice Hamiltonian with a two-fold degenerate $f$-level in each site, corresponding to $5f^2$ electronic configuration with $S=1$ spins. A derivation of the Schrieffer-Wolff transformation is presented and the resulting Hamiltonian has an effective $f$-band term, in addition to the regular exchange Kondo interaction between the $S=1$ $f$-spins and the $s=1/2$ spins of the conduction electrons. The obtained effective Kondo lattice model can describe both the Kondo regime and a weak delocalization of $5f$-electron. Within this model we compute the Kondo and Curie temperatures as a function of model parameters, namely the Kondo exchange interaction constant $J_K$, the magnetic intersite exchange interaction $J_H$ and the effective $f$-bandwidth. We deduce, therefore, a phase diagram of the model which yields the coexistence of Kondo effect and ferromagnetic ordering and also accounts for the pressure dependence of the Curie temperature of uranium compounds such as UTe.

cond-mat.str-el↗

arXiv2009

Exploring the Spectrum of QCD using the Lattice

The calculation of the spectrum of QCD is key to an understanding of the strong interactions, and vital if we are to capitalize on the experimental study of the spectrum. In this paper, we describe progress towards understanding the spectrum of resonances of both mesons and baryons from lattice QCD, focusing in particular on the resonances of the $I=1/2$ nucleon states, and of charmonium mesons composed of the heavy charmed quarks.

hep-lat↗

arXiv2009

The crucial role of CLEO-c in the measurement of gamma

The most sensitive method to measure the CKM angle gamma is to exploit interference in B(+/-)->DK(+/-) decays, with the D-meson decaying to a hadronic final state. The analysis of quantum-correlated decays of the psi(3770) at CLEO-c provides invaluable information on the strong-phase difference between the D0 and D0bar across the Dalitz plane. Results from analyses of the decays D->K0PiPi and D->K0KK will be presented.

hep-ex↗

arXiv2009

Exotic and excited-state radiative transitions in charmonium from lattice QCD

We compute, for the first time using lattice QCD methods, radiative transition rates involving excited charmonium states, states of high spin and exotics. Utilizing a large basis of interpolating fields we are able to project out various excited state contributions to three-point correlators computed on quenched anisotropic lattices. In the first lattice QCD calculation of the exotic 1-+ eta_c1 radiative decay, we find a large partial width Gamma(eta_c1 -> J/psi gamma) ~ 100 keV. We find clear signals for electric dipole and magnetic quadrupole transition form factors in chi_c2 -> J/psi gamma, calculated for the first time in this framework, and study transitions involving excited psi and chi_c1,2 states. We calculate hindered magnetic dipole transition widths without the sensitivity to assumptions made in model studies and find statistically significant signals, including a non-exotic vector hybrid candidate Y_hyb? -> eta_c gamma. As well as comparison to experimental data, we discuss in some detail the phenomenology suggested by our results and the extent to which it mirrors that of quark potential models and make suggestions for the interpretation of our results involving exotic quantum numbered states.

hep-ph↗