arXiv subjects

Pan He

Publications and source records attributed to Pan He.

47 records · Page 3Linked to original sources

arXiv2018

Adversarial Examples: Attacks and Defenses for Deep Learning

With rapid progress and significant successes in a wide spectrum of applications, deep learning is being applied in many safety-critical environments. However, deep neural networks have been recently found vulnerable to well-designed input samples, called adversarial examples. Adversarial examples are imperceptible to human but can easily fool deep neural networks in the testing/deploying stage. The vulnerability to adversarial examples becomes one of the major risks for applying deep neural networks in safety-critical environments. Therefore, attacks and defenses on adversarial examples draw great attention. In this paper, we review recent findings on adversarial examples for deep neural networks, summarize the methods for generating adversarial examples, and propose a taxonomy of these methods. Under the taxonomy, applications for adversarial examples are investigated. We further elaborate on countermeasures for adversarial examples and explore the challenges and the potential solutions.

cs.LG↗

arXiv2018

Observation of out-of-plane spin texture in a SrTiO3 (111) two-dimensional electron gas

We explore the second order bilinear magnetoelectric resistance (BMER) effect in the d-electron-based two-dimensional electron gas (2DEG) at the SrTiO3 (111) surface. We find an evidence of a spin-split band structure with the archetypal spin-momentum locking of the Rashba effect for the in-plane component. Under an out-of-plane magnetic field, we find a BMER signal that breaks the six-fold symmetry of the electronic dispersion, which is a fingerprint for the presence of a momentum dependent out-of-plane spin component. Relativistic electronic structure calculations reproduce this spin-texture and indicate that the out-of-plane component is a ubiquitous property of oxide 2DEGs arising from strong crystal field effects. We further show that the BMER response of the SrTiO3 (111) 2DEG is tunable and unexpectedly large.

cond-mat.mes-hall↗

arXiv2018

Direct visualization of current induced spin accumulation in topological insulators

Charge-to-spin conversion in various material systems is the key for the fundamental understanding of spin-orbitronis as well as the development of efficient means to manipulate the magnetization. We report the direct spatial imaging of current induced spin accumulation at the channel edges of Bi2Se3 and BiSbTeSe2 topological insulators by a scanning photovoltage microscope at room temperature. The spin polarization is along the out-of-plane direction with opposite signs for the two channel edges. The accumulated spin direction reverses sign upon changing the current direction and the detected spin signal shows a linear dependence on the magnitude of currents, which indicates that our observed phenomena are the current induced effects. The spin Hall angle of Bi2Se3 and BiSbTeSe2 is determined to be 0.0085 and 0.0616, respectively. We further image the current induced spin accumulation in a Pt heavy metal. Our results open up the possibility of optically detecting the current induced spin accumulations, and thus point towards a better understanding of the interaction between spins and circularly polarized light.

cond-mat.mes-hall↗

arXiv2018

Boosting up Scene Text Detectors with Guided CNN

Deep CNNs have achieved great success in text detection. Most of existing methods attempt to improve accuracy with sophisticated network design, while paying less attention on speed. In this paper, we propose a general framework for text detection called Guided CNN to achieve the two goals simultaneously. The proposed model consists of one guidance subnetwork, where a guidance mask is learned from the input image itself, and one primary text detector, where every convolution and non-linear operation are conducted only in the guidance mask. On the one hand, the guidance subnetwork filters out non-text regions coarsely, greatly reduces the computation complexity. On the other hand, the primary text detector focuses on distinguishing between text and hard non-text regions and regressing text bounding boxes, achieves a better detection accuracy. A training strategy, called background-aware block-wise random synthesis, is proposed to further boost up the performance. We demonstrate that the proposed Guided CNN is not only effective but also efficient with two state-of-the-art methods, CTPN and EAST, as backbones. On the challenging benchmark ICDAR 2013, it speeds up CTPN by 2.9 times on average, while improving the F-measure by 1.5%. On ICDAR 2015, it speeds up EAST by 2.0 times while improving the F-measure by 1.0%.

cs.CV↗

arXiv2017

Single Shot Text Detector with Regional Attention

We present a novel single-shot text detector that directly outputs word-level bounding boxes in a natural image. We propose an attention mechanism which roughly identifies text regions via an automatically learned attentional map. This substantially suppresses background interference in the convolutional features, which is the key to producing accurate inference of words, particularly at extremely small sizes. This results in a single model that essentially works in a coarse-to-fine manner. It departs from recent FCN- based text detectors which cascade multiple FCN models to achieve an accurate prediction. Furthermore, we develop a hierarchical inception module which efficiently aggregates multi-scale inception features. This enhances local details, and also encodes strong context information, allow- ing the detector to work reliably on multi-scale and multi- orientation text with single-scale images. Our text detector achieves an F-measure of 77% on the ICDAR 2015 bench- mark, advancing the state-of-the-art results in [18, 28]. Demo is available at: http://sstd.whuang.org/.

cs.CV↗

arXiv2017

Bilinear magneto-electric resistance as a probe of three-dimensional spin texture in topological surface states

Surface states of three-dimensional topological insulators exhibit the phenomenon of spin-momentum locking, whereby the orientation of an electron spin is determined by its momentum. Probing the spin texture of these states is of critical importance for the realization of topological insulator devices, however the main technique available so far is the spin- and angle-resolved photoemission spectroscopy. Here we reveal a close link between the spin texture and a new kind of magneto-resistance, which depends on the relative orientation of the current with respect to the magnetic field as well as the crystallographic axes, and scales linearly with both the applied electric and magnetic fields. This bilinear magneto-electric resistance can be used to map the spin texture of topological surface states by simple transport measurements. For a prototypical Bi2Se3 single layer, we can map both the in-plane and the out-of-plane components of the spin texture - the latter arising from hexagonal warping. Theoretical calculations suggest that the bilinear magneto-electric resistance originates from the conversion of a non-equilibrium spin current into a charge current under the application of the external magnetic field.

cond-mat.mes-hall↗

arXiv2017

Continuous Tuning the Magnitude and Direction of Spin-Orbit Torque Using Bilayer Heavy Metals

Spin-orbit torques (SOTs) have opened a new path to switch the magnetization in perpendicularly magnetized films and are of great interest due to their potential applications in novel data storage technology, such as the magnetic random access memory (MRAM). The effective manipulation of SOT has thus become an important step towards these applications. Here, current induced spin-orbit effective fields and magnetization switching are investigated in Pt/Ta/CoFeB/MgO structures with bilayer heavy metals. With a fixed thickness (1 nm) of the Ta layer, the magnitude and sign of current induced spin-orbit effective fields can be continuously tuned by changing the Pt layer thickness, consistent with the current induced magnetization switching data. The ratio of longitudinal to transverse spin-orbit effective fields is found to be determined by the Ta/CoFeB interface and can be continuously tuned by changing the Pt layer thickness. The Dzyaloshinskii-Moriya interaction (DMI) is found to be weak and shows an insignificant variation with the Pt thickness. The results demonstrate an effective method to tune SOTs utilizing bilayer heavy metals without affecting the DMI, a desirable feature which will be useful for the design of SOT-based devices.

cond-mat.mtrl-sci↗

arXiv2016

Enhanced spin-orbit torque via modulation of spin current absorption

The magnitude of spin-orbit torque (SOT), exerted to a ferromagnet (FM) from an adjacent heavy metal (HM), strongly depends on the amount of spin currents absorbed in the FM. We exploit the large spin absorption at the Ru interface to manipulate the SOTs in HM/FM/Ru multilayers. While the FM thickness is smaller than its spin dephasing length of 1.2 nm, the top Ru layer largely boosts the absorption of spin currents into the FM layer and substantially enhances the strength of SOT acting on the FM. Spin-pumping experiments induced by ferromagnetic resonance support our conclusions that the observed increase in the SOT efficiency can be attributed to an enhancement of the spin-current absorption. A theoretical model that considers both reflected and transmitted mixing conductances at the two interfaces of FM is developed to explain the results.

cond-mat.mtrl-sci↗

arXiv2016

Detecting Text in Natural Image with Connectionist Text Proposal Network

We propose a novel Connectionist Text Proposal Network (CTPN) that accurately localizes text lines in natural image. The CTPN detects a text line in a sequence of fine-scale text proposals directly in convolutional feature maps. We develop a vertical anchor mechanism that jointly predicts location and text/non-text score of each fixed-width proposal, considerably improving localization accuracy. The sequential proposals are naturally connected by a recurrent neural network, which is seamlessly incorporated into the convolutional network, resulting in an end-to-end trainable model. This allows the CTPN to explore rich context information of image, making it powerful to detect extremely ambiguous text. The CTPN works reliably on multi-scale and multi- language text without further post-processing, departing from previous bottom-up methods requiring multi-step post-processing. It achieves 0.88 and 0.61 F-measure on the ICDAR 2013 and 2015 benchmarks, surpass- ing recent results [8, 35] by a large margin. The CTPN is computationally efficient with 0:14s/image, by using the very deep VGG16 model [27]. Online demo is available at: http://textdet.com/.

cs.CV↗

arXiv2015

Reading Scene Text in Deep Convolutional Sequences

We develop a Deep-Text Recurrent Network (DTRN) that regards scene text reading as a sequence labelling problem. We leverage recent advances of deep convolutional neural networks to generate an ordered high-level sequence from a whole word image, avoiding the difficult character segmentation problem. Then a deep recurrent model, building on long short-term memory (LSTM), is developed to robustly recognize the generated CNN sequences, departing from most existing approaches recognising each character independently. Our model has a number of appealing properties in comparison to existing scene text recognition methods: (i) It can recognise highly ambiguous words by leveraging meaningful context information, allowing it to work reliably without either pre- or post-processing; (ii) the deep CNN feature is robust to various image distortions; (iii) it retains the explicit order information in word image, which is essential to discriminate word strings; (iv) the model does not depend on pre-defined dictionary, and it can process unknown words and arbitrary strings. Codes for the DTRN will be available.

cs.CV↗

arXiv2012

Scaling of intrinsic Gilbert damping with spin-orbital coupling strength

We have experimentally and theoretically investigated the dependence of the intrinsic Gilbert damping parameter $α_0$ on the spin-orbital coupling strength $ξ$ by using L1$_{\mathrm{0}}$ ordered FePd$_{\mathrm{1-x}}$Pt$_{\mathrm{x}}$ ternary alloy films with perpendicular magnetic anisotropy. With the time-resolved magneto-optical Kerr effect, $α_0$ is found to increase by more than a factor of ten when $x$ varies from 0 to 1.0. Since changes of other leading parameters are found to be neglected, the $α_0$ has for the first time been proven to be proportional to $ξ^2$.

cond-mat.mtrl-sci↗