arXiv subjects

Anil Jain

Publications and source records attributed to Anil Jain.

30 records · Page 2Linked to original sources

arXiv2023

Cluster and Aggregate: Face Recognition with Large Probe Set

Feature fusion plays a crucial role in unconstrained face recognition where inputs (probes) comprise of a set of $N$ low quality images whose individual qualities vary. Advances in attention and recurrent modules have led to feature fusion that can model the relationship among the images in the input set. However, attention mechanisms cannot scale to large $N$ due to their quadratic complexity and recurrent modules suffer from input order sensitivity. We propose a two-stage feature fusion paradigm, Cluster and Aggregate, that can both scale to large $N$ and maintain the ability to perform sequential inference with order invariance. Specifically, Cluster stage is a linear assignment of $N$ inputs to $M$ global cluster centers, and Aggregation stage is a fusion over $M$ clustered features. The clustered features play an integral role when the inputs are sequential as they can serve as a summarization of past features. By leveraging the order-invariance of incremental averaging operation, we design an update rule that achieves batch-order invariance, which guarantees that the contributions of early image in the sequence do not diminish as time steps increase. Experiments on IJB-B and IJB-S benchmark datasets show the superiority of the proposed two-stage paradigm in unconstrained face recognition. Code and pretrained models are available in https://github.com/mk-minchul/caface

cs.CV↗

arXiv2022

Controllable and Guided Face Synthesis for Unconstrained Face Recognition

Although significant advances have been made in face recognition (FR), FR in unconstrained environments remains challenging due to the domain gap between the semi-constrained training datasets and unconstrained testing scenarios. To address this problem, we propose a controllable face synthesis model (CFSM) that can mimic the distribution of target datasets in a style latent space. CFSM learns a linear subspace with orthogonal bases in the style latent space with precise control over the diversity and degree of synthesis. Furthermore, the pre-trained synthesis model can be guided by the FR model, making the resulting images more beneficial for FR model training. Besides, target dataset distributions are characterized by the learned orthogonal bases, which can be utilized to measure the distributional similarity among face datasets. Our approach yields significant performance gains on unconstrained benchmarks, such as IJB-B, IJB-C, TinyFace and IJB-S (+5.76% Rank1).

cs.CV↗

arXiv2022

Observation of anisotropic thermal expansion and Jahn-Teller effect in double perovskites Sr$_{2-x}$La$_x$CoNbO$_6$ using neutron diffraction

We use temperature dependent neutron powder diffraction (NPD) to investigate the structural changes and magnetic interactions in double perovskite Sr$_{2-x}$La$_x$CoNbO$_6$ ($x=$ 0.4 and 0.6). A structural phase transition from tetragonal ($I4/m$) to monoclinic ($P2_1/n$) is observed between $x=$ 0.4 and 0.6 samples. Interestingly, temperature evolution of the unit cell parameters follow the Grüneisen approximation, and the analysis suggest the isotropic thermal expansion in case of the $x=$ 0.4 sample, whereas the $x=$ 0.6 sample shows the anisotropy where the thermal expansion along the $c$-axis significantly deviates from the Grüneisen function. We observe the $z$-out Jahn-Teller distortion in the CoO$_6$ and consequently $z$-in distortion in the adjacent NbO$_6$ octahedra. With increase in the La substitution, a decrease in the degree of octahedral distortion is evident from the significant reduction in the local distortion parameter $Δ$ around the B-site atoms.

cond-mat.str-el↗

arXiv2021

Towards the Memorization Effect of Neural Networks in Adversarial Training

Recent studies suggest that ``memorization'' is one important factor for overparameterized deep neural networks (DNNs) to achieve optimal performance. Specifically, the perfectly fitted DNNs can memorize the labels of many atypical samples, generalize their memorization to correctly classify test atypical samples and enjoy better test performance. While, DNNs which are optimized via adversarial training algorithms can also achieve perfect training performance by memorizing the labels of atypical samples, as well as the adversarially perturbed atypical samples. However, adversarially trained models always suffer from poor generalization, with both relatively low clean accuracy and robustness on the test set. In this work, we study the effect of memorization in adversarial trained DNNs and disclose two important findings: (a) Memorizing atypical samples is only effective to improve DNN's accuracy on clean atypical samples, but hardly improve their adversarial robustness and (b) Memorizing certain atypical samples will even hurt the DNN's performance on typical samples. Based on these two findings, we propose Benign Adversarial Training (BAT) which can facilitate adversarial training to avoid fitting ``harmful'' atypical samples and fit as more ``benign'' atypical samples as possible. In our experiments, we validate the effectiveness of BAT, and show it can achieve better clean accuracy vs. robustness trade-off than baseline methods, in benchmark datasets such as CIFAR100 and Tiny~ImageNet.

cs.LG↗

arXiv2020

Experimental realisation of multipartite entanglement via quantum Fisher information in a uniform antiferromagnetic quantum spin chain

Quantum entanglement is a quantum mechanical phenomenon where the quantum state of a many-body system with many degrees of freedom cannot be described independently of the state of each body with a given degree of freedom, no matter how far apart in space each body is. Entanglement is not only considered a resource in quantum information but also believed to affect complex condensed matter systems. Detecting and quantifying multi-particle entanglement in a many-body system is thus of fundamental significance for both quantum information science and condensed matter physics. Here, we detect and quantify multipartite entanglement in a spin 1/2 Heisenberg antiferromagnetic chain in a bulk solid. Multipartite entanglement was detected using quantum Fisher information which was obtained using dynamic susceptibility measured via inelastic neutron scattering. The scaling behaviour of quantum Fisher information was found to identify the spin 1/2 Heisenberg antiferromagnetic chain to belong to a class of strongly entangled quantum phase transitions with divergent multipartite entanglement.

quant-ph↗

arXiv2020

On the Detection of Digital Face Manipulation

Detecting manipulated facial images and videos is an increasingly important topic in digital media forensics. As advanced face synthesis and manipulation methods are made available, new types of fake face representations are being created which have raised significant concerns for their use in social media. Hence, it is crucial to detect manipulated face images and localize manipulated regions. Instead of simply using multi-task learning to simultaneously detect manipulated images and predict the manipulated mask (regions), we propose to utilize an attention mechanism to process and improve the feature maps for the classification task. The learned attention maps highlight the informative regions to further improve the binary classification (genuine face v. fake face), and also visualize the manipulated regions. To enable our study of manipulated face detection and localization, we collect a large-scale database that contains numerous types of facial forgeries. With this dataset, we perform a thorough analysis of data-driven fake face detection. We show that the use of an attention mechanism improves facial forgery detection and manipulated region localization.

cs.CV↗

arXiv2019

Spin dynamics and unconventional magnetism in insulating La$_{(1-2x)}$Sr$_{2x}$Co$_{(1-x)}$Nb$_{x}$O$_3$

We study the structural, magnetic, transport and electronic properties of LaCoO$_3$ with Sr/Nb co-substitution, i.e., La$_{(1-2x)}$Sr$_{2x}$Co$_{(1-x)}$Nb$_{x}$O$_3$ using x-ray and neutron diffraction, dc and ac-magnetization, neutron depolarization, dc-resistivity and photoemission measurements. The powder x-ray and neutron diffraction data were fitted well with the rhombohedral crystal symmetry (space group \textit{R$\bar{3}$c}) in Rietveld refinement analysis. The calculated effective magnetic moment ($\approx$3.85~$μ_B$) and average spin ($\approx$1.5) of Co ions from the analysis of magnetic susceptibility data are consistent with 3+ state of Co ions in intermediate-spin (IS) and high-spin (HS) states in the ratio of $\approx$50:50, i.e., spin-state of Co$^{3+}$ is preserved at least up to $x=$ 0.1 sample. Interestingly, the magnetization values were significantly increased with respect to the $x=$ 0 sample, and the M-H curves show non-saturated behavior up to an applied maximum magnetic field of $\pm$70 kOe. The ac-susceptibility data show a shift in the freezing temperature with excitation frequency and the detailed analysis confirm the slower dynamics and a non-zero value of the Vogel-Fulcher temperature T$_0$, which suggests for the cluster spin glass. The unusual magnetic behavior indicates the presence of complex magnetic interactions at low temperatures. The dc-resistivity measurements show the insulating nature in all the samples. However, relatively large density of states $\approx$10$^{22}$ eV$^{-1}$cm$^{-3}$ and low activation energy $\approx$130~meV are found in $x=$ 0.05 sample. Using x-ray photoemission spectroscopy, we study the core-level spectra of La 3$d$, Co 2$p$, Sr 3$d$, and Nb 3$d$ to confirm the valence state.

cond-mat.str-el↗

arXiv2018

Face Recognition: Primates in the Wild

We present a new method of primate face recognition, and evaluate this method on several endangered primates, including golden monkeys, lemurs, and chimpanzees. The three datasets contain a total of 11,637 images of 280 individual primates from 14 species. Primate face recognition performance is evaluated using two existing state-of-the-art open-source systems, (i) FaceNet and (ii) SphereFace, (iii) a lemur face recognition system from literature, and (iv) our new convolutional neural network (CNN) architecture called PrimNet. Three recognition scenarios are considered: verification (1:1 comparison), and both open-set and closed-set identification (1:N search). We demonstrate that PrimNet outperforms all of the other systems in all three scenarios for all primate species tested. Finally, we implement an Android application of this recognition system to assist primate researchers and conservationists in the wild for individual recognition of primates.

cs.CV↗

arXiv2017

Raman scattering from Higgs mode oscillations in the two-dimensional antiferromagnet Ca$_2$RuO$_4$

We present and analyze Raman spectra of the Mott insulator Ca$_2$RuO$_4$, whose quasi-two-dimensional antiferromagnetic order has been described as a condensate of low-lying spin-orbit excitons with angular momentum $J_{eff}=1$. In the $A_g$ polarization geometry, the amplitude (Higgs) mode of the spin-orbit condensate is directly probed in the scalar channel, thus avoiding infrared-singular magnon contributions. In the $B_{1g}$ geometry, we observe a single-magnon peak as well as two-magnon and two-Higgs excitations. Model calculations using exact diagonalization quantitatively agree with the observations. Together with recent neutron scattering data, our study provides strong evidence for excitonic magnetism in Ca$_2$RuO$_4$ and points out new perspectives for research on the Higgs mode in two dimensions.

cond-mat.str-el↗

arXiv2015

Similarity Learning via Adaptive Regression and Its Application to Image Retrieval

We study the problem of similarity learning and its application to image retrieval with large-scale data. The similarity between pairs of images can be measured by the distances between their high dimensional representations, and the problem of learning the appropriate similarity is often addressed by distance metric learning. However, distance metric learning requires the learned metric to be a PSD matrix, which is computational expensive and not necessary for retrieval ranking problem. On the other hand, the bilinear model is shown to be more flexible for large-scale image retrieval task, hence, we adopt it to learn a matrix for estimating pairwise similarities under the regression framework. By adaptively updating the target matrix in regression, we can mimic the hinge loss, which is more appropriate for similarity learning problem. Although the regression problem can have the closed-form solution, the computational cost can be very expensive. The computational challenges come from two aspects: the number of images can be very large and image features have high dimensionality. We address the first challenge by compressing the data by a randomized algorithm with the theoretical guarantee. For the high dimensional issue, we address it by taking low rank assumption and applying alternating method to obtain the partial matrix, which has a global optimal solution. Empirical studies on real world image datasets (i.e., Caltech and ImageNet) demonstrate the effectiveness and efficiency of the proposed method.

cs.LG↗

arXiv2014

Competing exchange interactions on the verge of a metal-insulator transition in the two-dimensional spiral magnet Sr$_3$Fe$_2$O$_7$

We report a neutron scattering study of the magnetic order and dynamics of the bilayer perovskite Sr$_3$Fe$_2$O$_7$, which exhibits a temperature-driven metal-insulator transition at 340 K. We show that the Fe$^{4+}$ moments adopt incommensurate spiral order below $T_\text{N}=115$ K and provide a comprehensive description of the corresponding spin wave excitations. The observed magnetic order and excitation spectra can be well understood in terms of an effective spin Hamiltonian with interactions ranging up to third nearest-neighbor pairs. The results indicate that the helical magnetism in Sr$_3$Fe$_2$O$_7$ results from competition between ferromagnetic double-exchange and antiferromagnetic superexchange interactions whose strengths become comparable near the metal-insulator transition. They thus confirm a decades-old theoretical prediction and provide a firm experimental basis for models of magnetic correlations in strongly correlated metals.

cond-mat.str-el↗

arXiv2013

One-Dimensional Dispersive Magnon Excitation in the Frustrated Spin-2 Chain System Ca3Co2O6

Using inelastic neutron scattering, we have observed a quasi-one-dimensional dispersive magnetic excitation in the frustrated triangular-lattice spin-2 chain oxide Ca3Co2O6. At the lowest temperature (T = 1.5 K), this magnon is characterized by a large zone-center spin gap of ~27 meV, which we attribute to the large single-ion anisotropy, and disperses along the chain direction with a bandwidth of ~3.5 meV. In the directions orthogonal to the chains, no measurable dispersion was found. With increasing temperature, the magnon dispersion shifts towards lower energies, yet persists up to at least 150 K, indicating that the ferromagnetic intrachain correlations survive up to 6 times higher temperatures than the long-range interchain antiferromagnetic order. The magnon dispersion can be well described within the predictions of linear spin-wave theory for a system of weakly coupled ferromagnetic chains with large single-ion anisotropy, enabling the direct quantitative determination of the magnetic exchange and anisotropy parameters.

cond-mat.str-el↗