Showing papers by "Panasonic published in 2018"

Scale-Aware Fast R-CNN for Pedestrian Detection

Abstract: In this paper, we consider the problem of pedestrian detection in natural scenes. Intuitively, instances of pedestrians with different spatial scales may exhibit dramatically different features. Thus, large variance in instance scales, which results in undesirable large intracategory variance in features, may severely hurt the performance of modern object instance detection methods. We argue that this issue can be substantially alleviated by the divide-and-conquer philosophy. Taking pedestrian detection as an example, we illustrate how we can leverage this philosophy to develop a Scale-Aware Fast R-CNN (SAF R-CNN) framework. The model introduces multiple built-in subnetworks which detect pedestrians with scales from disjoint ranges. Outputs from all of the subnetworks are then adaptively combined to generate the final detection results that are shown to be robust to large variance in instance scales, via a gate function defined over the sizes of object proposals. Extensive evaluations on several challenging pedestrian detection datasets well demonstrate the effectiveness of the proposed SAF R-CNN. Particularly, our method achieves state-of-the-art performance on Caltech [P. Dollar, C. Wojek, B. Schiele, and P. Perona, “Pedestrian detection: An evaluation of the state of the art,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 34, no. 4, pp. 743–761, Apr. 2012], and obtains competitive results on INRIA [N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2005, pp. 886–893], ETH [A. Ess, B. Leibe, and L. V. Gool, “Depth and appearance for mobile scene analysis,” in Proc. Int. Conf. Comput. Vis ., 2007, pp. 1–8], and KITTI [A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? The KITTI vision benchmark suite,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit ., 2012, pp. 3354–3361].

Solid Halide Electrolytes with High Lithium-Ion Conductivity for Application in 4 V Class Bulk-Type All-Solid-State Batteries.

Abstract: New lithium halide solid-electrolyte materials, Li3 YCl6 and Li3 YBr6 , are found to exhibit high lithium-ion conductivity, high deformability, and high chemical and electrochemical stability, which are required properties for all-solid-state battery (ASSB) applications, particularly for large-scale deployment. The lithium-ion conductivities of cold-pressed powders surpass 1 mS cm-1 at room temperature without additional intergrain or grain boundary resistances. Bulk-type ASSB cells employing these new halide solid electrolyte materials exhibit coulombic efficiencies as high as 94% with an active cathode material of LiCoO2 without any extra coating. These superior electrochemical characteristics, as well as their material stability, indicate that lithium halide salts are another promising candidate for ASSB solid electrolytes in addition to sulfides or oxides.

Grain boundary dominated charge transport in Mg3Sb2-based compounds

Abstract: Thermally activated mobility near room temperature is a signature of detrimental scattering that limits the efficiency and figure-of-merit zT in thermoelectric semiconductors. This effect has been observed dramatically in Mg3Sb2-based compounds, but also to a lesser extent in other thermoelectric compounds. Processing samples differently or adding impurities such that this effect is less noticeable produces materials with a higher zT. Experiments suggest that the behavior is related to grain boundaries, but impurity scattering has also been proposed. However, conventional models using Matthissen's rule are not able to explain the dramatic change in the temperature dependency of conductivity or drift mobility which is observed in Mg3Sb2-based compounds. We find that it is essential to consider the grain boundary region as an effectively separate phase rather than a scattering center, taking into account the weaker screening in semiconductors compared with classical metals. By modeling a grain boundary phase with a band offset, we successfully reproduce the experimentally observed conductivity versus temperature and thermopower versus conductivity relations, which indicate an improved description of transport. The model shows good agreement with measured grain size dependencies of conductivity, opening up avenues for quantitatively engineering materials with similar behavior. Model estimates predict room for >60% improvement in the room temperature zT of Mg3.2Sb1.5Bi0.49Te0.01 if the grain boundary resistance could be eliminated.

Towards Pose Invariant Face Recognition in the Wild

Abstract: Pose variation is one key challenge in face recognition. As opposed to current techniques for pose invariant face recognition, which either directly extract pose invariant features for recognition, or first normalize profile face images to frontal pose before feature extraction, we argue that it is more desirable to perform both tasks jointly to allow them to benefit from each other. To this end, we propose a Pose Invariant Model (PIM) for face recognition in the wild, with three distinct novelties. First, PIM is a novel and unified deep architecture, containing a Face Frontalization sub-Net (FFN) and a Discriminative Learning sub-Net (DLN), which are jointly learned from end to end. Second, FFN is a well-designed dual-path Generative Adversarial Network (GAN) which simultaneously perceives global structures and local details, incorporated with an unsupervised cross-domain adversarial training and a "learning to learn" strategy for high-fidelity and identity-preserving frontal view synthesis. Third, DLN is a generic Convolutional Neural Network (CNN) for face recognition with our enforced cross-entropy optimization strategy for learning discriminative yet generalized feature representation. Qualitative and quantitative experiments on both controlled and in-the-wild benchmarks demonstrate the superiority of the proposed model over the state-of-the-arts.

Video-Based Person Re-Identification With Accumulative Motion Context

Abstract: Video-based person re-identification plays a central role in realistic security and video surveillance. In this paper, we propose a novel accumulative motion context (AMOC) network for addressing this important problem, which effectively exploits the long-range motion context for robustly identifying the same person under challenging conditions. Given a video sequence of the same or different persons, the proposed AMOC network jointly learns appearance representation and motion context from a collection of adjacent frames using a two-stream convolutional architecture. Then, AMOC accumulates clues from motion context by recurrent aggregation, allowing effective information flow among adjacent frames and capturing dynamic gist of the persons. The architecture of AMOC is end-to-end trainable, and thus, motion context can be adapted to complement appearance clues under unfavorable conditions ( e.g. , occlusions). Extensive experiments are conduced on three public benchmark data sets, i.e. , the iLIDS-VID, PRID-2011, and MARS data sets, to investigate the performance of AMOC. The experimental results demonstrate that the proposed AMOC network outperforms state-of-the-arts for video-based re-identification significantly and confirm the advantage of exploiting long-range motion context for video-based person re-identification, validating our motivation evidently.

A 4M Synapses integrated Analog ReRAM based 66.5 TOPS/W Neural-Network Processor with Cell Current Controlled Writing and Flexible Network Architecture

Abstract: This paper presents low-power neural-network (NN) processor using ReRAM to store weights as analog resistance for future AI computing. We propose ReRAM perceptron circuit for realizing large scale integration, highly accurate cell current controlled writing scheme, and flexible network architecture (FNA) in which any NNs can be configured. Fabricated 180nm test chip shows well-controlled analog cell current with linear 30μA dynamic range and 0.59μA variation of 1 sigma, results in 90.8% MNIST numerical recognition rate. Furthermore, 4M synapses integrated 40nm test chip achieves lower analog cell current and 66.5 TOPS/W power efficiency.

Enhancement of average thermoelectric figure of merit by increasing the grain-size of Mg3.2Sb1.5Bi0.49Te0.01

Abstract: Zintl compound n-type Mg3(Sb,Bi)2 was recently found to exhibit excellent thermoelectric figure of merit zT (∼1.5 at around 700 K). To improve the thermoelectric performance in the whole temperature range of operation from room temperature to 720 K, we investigated how the grain size of sintered samples influences electronic and thermal transport. By increasing the average grain size from 1.0 μm to 7.8 μm, the Hall mobility below 500 K was significantly improved, possibly due to suppression of grain boundary scattering. We also confirmed that the thermal conductivity did not change by increasing the grain size. Consequently, the sample with larger grains exhibited enhanced average zT. The calculated efficiency of thermoelectric power generation reaches 14.5% (ΔT = 420 K), which is quite high for a polycrystalline pristine material.

Boosting the Efficiency of Perovskite Solar Cells with CsBr-Modified Mesoporous TiO2 Beads as Electron-Selective Contact

Abstract: Rapid extraction of photogenerated charge carriers is essential to achieve high efficiencies with perovskite solar cells (PSCs). Here, a new mesoscopic architecture as electron-selective contact for PSCs featuring 40 nm sized TiO2 beads endowed with mesopores of a few nanometer diameters is introduced. The bimodal pore distribution inherent to these films produces a very large contact area of 200 m2 g−1 whose access by the perovskite light absorber is facilitated by the interstitial voids between the particles. Modification of the TiO2 surface by CsBr further strengthens its interaction with the perovskite. As a result, photogenerated electrons are extracted rapidly producing a very high fill factor of close to 80% a VOC of 1.14 V and a PCE up to 21% with negligible hysteresis.

Efficacy and safety of nivolumab in non-small cell lung cancer with preexisting interstitial lung disease.

Abstract: Background The risk of developing lung cancer is high in patients with interstitial lung disease (ILD), as few treatment options are available. Immune checkpoint inhibitors (ICI) are used for the treatment of non-small cell lung cancer (NSCLC) in clinical practice; however, in patients with preexisting ILD, the risk of ICI-related pneumonitis is unknown. We evaluated the efficacy and lung toxicity of nivolumab in patients with NSCLC and ILD. Methods We retrospectively reviewed the medical records of 216 NSCLC patients who had received nivolumab therapy. The existence of ILD in these patients was determined by lung computed tomography findings; 26 patients had ILD. We evaluated the efficacy of nivolumab by measuring the response rate (RR), progression-free survival (PFS) duration, and lung toxicity by incidence, severity, and outcome of nivolumab-related ILD. Results The RR and median PFS of the ILD and non-ILD groups were 27% versus 13% (P = 0.078) and 2.7 (95% confidence interval [CI], 1.7-5.3) versus 2.9 months (95% CI 2.1-3.4; P = 0.919), respectively. The incidences of total and severe nivolumab-related pneumonitis were significantly higher in the ILD group than in the non-ILD group (31% vs. 12%, P = 0.014 and 19% vs. 5%, P = 0.022, respectively). No death from nivolumab-related pneumonitis occurred. Over 50% of the patients in both groups with nivolumab-related pneumonitis showed improvement over time. Conclusion Relative to the non-ILD group, nivolumab-related pneumonitis was observed more frequently in the ILD group; however, most cases were manageable.

Effects of Electrolyte Buffer Capacity on Surface Reactant Species and the Reaction Rate of CO2 in Electrochemical CO2 Reduction

Abstract: In the aqueous electrochemical reduction of CO2, the choice of electrolyte is responsible for the catalytic activity and selectivity, although there remains a need for more in-depth understanding of electrolyte effects and mechanisms. In this study, using both experimental and simulation approaches, we report how the buffer capacity of the electrolytes affects the kinetics and equilibrium of surface reactant species and the resulting reaction rate of CO2 with varying partial CO2 pressure. Electrolytes investigated include KCl (nonbuffered), KHCO3 (buffered by bicarbonate), and phosphate-buffered electrolytes. Assuming 100% methane production, the simulation successfully explains the experimental trends in maximum CO2 flux in KCl and KHCO3 and also highlights the difference between KHCO3 and phosphate in terms of pKa as well as the impact of the buffer capacity. To examine the electrolyte impact on selectivity, the model is run with a constant total current density. Using this model, several factors are el...

3D-Aided Deep Pose-Invariant Face Recognition.

Abstract: Learning from synthetic faces, though perhaps appealing for high data efficiency, may not bring satisfactory performance due to the distribution discrepancy of the synthetic and real face images. To mitigate this gap, we propose a 3D-Aided Deep Pose-Invariant Face Recognition Model (3D-PIM), which automatically recovers realistic frontal faces from arbitrary poses through a 3D face model in a novel way. Specifically, 3D-PIM incorporates a simulator with the aid of a 3D Morphable Model (3D MM) to obtain shape and appearance prior for accelerating face normalization learning, requiring less training data. It further leverages a global-local Generative Adversarial Network (GAN) with multiple critical improvements as a refiner to enhance the realism of both global structures and local details of the face simulator’s output using unlabelled real data only, while preserving the identity information. Qualitative and quantitative experiments on both controlled and in-the-wild benchmarks clearly demonstrate superiority of the proposed model over state-of-the-arts.

Development and evaluation of a palm-sized optical PM2.5 sensor

Abstract: A new palm-sized optical PM2.5 sensor has been developed and its performance evaluated. The PM2.5 mass concentration was calculated from the distribution of light scattering intensity by considerin...

Radiated EMI Estimation From DC–DC Converters With Attached Cables Based on Terminal Equivalent Circuit Modeling

Abstract: An equivalent two terminal model based on the Thevenin equivalents describes the common mode (CM) currents on the input and output side of two widely used types of dc-to-dc power converters—buck and boost. Thus, it describes a nonlinear circuit by a linear equivalent circuit. The maximized spectrum of the CM currents is extracted for converters with stochastic signals using a novel characterization procedure. The extracted Thevenin model is used in co-simulation combined with a full-wave electromagnetic solver to predict the radiated emissions from the system consisting of the shielded dc–dc converters with attached cables and a DC brushless motor as load. The results using the terminal model agree well with the measurements providing that the actual CM loads are within the range of CM loads used while obtaining the Thevenin equivalent circuit.

Multiple prose group communication during a sidelink control period

Abstract: A method allocates radio resources for a transmitting user equipment (UE) to perform direct communication transmissions over a direct sidelink connection to one or more receiving UEs. At least two sidelink grant processes are provided in the transmitting UE for the transmitting UE to handle at least two sidelink grants within the same transmission control period. Each one of the sidelink grant processes is associated with an identification and can be associated with one sidelink grant. The transmitting UE acquires at least two sidelink grants and associates each of the acquired sidelink grants with one sidelink grant process. For each sidelink grant, the transmitting UE allocates radio resources according to the respective sidelink grant to perform a direct communication transmission of sidelink control information and of data over the direct sidelink connection. The transmitting UE performs a direct communication transmission per acquired sidelink grant within the same transmission control period.

Quantitative Operando Visualization of Electrochemical Reactions and Li Ions in All-Solid-State Batteries by STEM-EELS with Hyperspectral Image Analyses.

Abstract: All-solid-state lithium-ion batteries (LIBs) are one of the promising candidates to overcome some issues of conventional LIBs with liquid electrolytes. However, high interfacial resistance of Li-ion transfer at the electrode/solid electrolyte limits their performance. Thus, it is important to clarify interfacial phenomena in a nanometer scale. Here, we present a new method to dynamically observe the Li-ion distribution and Co-ion electronic states in a LiCoO2 cathode of the all-solid-state LIB during charge and discharge reactions using operando scanning transmission electron microscopy (STEM) and electron energy-loss spectroscopy (EELS). By applying a hyperspectral image analysis of non-negative matrix factorization (NMF) to the STEM-EELS, we succeeded in clearly observing the quantitative Li-ion distribution in the operando condition. We found from the operando observation with NMF that the Li ions did not uniformly extract/insert during the charge/discharge reactions, and the activity of the electroche...

Energy efficiency of ionic transport through proton conducting ceramic electrolytes for energy conversion applications

Abstract: In this study, the efficiency of ionic transport through proton-conducting ceramic electrolytes for energy conversion applications was discussed while considering the energy losses caused by the mixed ionic and electronic conduction and by the ionic transport resistance. It was shown that high energy efficiency cannot be expected with an extremely thin electrolyte because of the significant energy loss due to the internal current leakage. In the case of a fuel cell with a BaZr0.8Y0.2O3−δ electrolyte operating at 873 K, the maximum energy efficiency was estimated to be only 82% when the product of the external current density and the electrolyte thickness was 1.7 × 10−3 A cm−1. The energy efficiency for the ionic transport in BaZr0.8Y0.2O3−δ was compared with those for typical oxide ion conductors, i.e. yttria stabilized zirconia and Ce0.9Gd0.1O1.95−δ, under various fuel cell operating conditions. BaZr0.8Y0.2O3−δ can provide higher energy efficiency than the oxide ion conductors at lower temperatures and under higher current conditions. For the efficient operation of fuel cells using a proton-conducting ceramic as an electrolyte, lower operating temperatures below 873 K is preferable from the viewpoint of energy loss in the electrolyte. The influence of the partial conductivity variation on the energy efficiency was also discussed. It was shown that not only the increase of ionic conductivity but also the decrease of electronic conductivity could improve the energy efficiency if the electrolyte thickness can be optimized. This means that decreasing electronic conductivity as well as increasing proton conductivity can be an effective strategy to develop highly efficient proton conductors for energy conversion applications.

Dual-Mode Vehicle Motion Pattern Learning for High Performance Road Traffic Anomaly Detection

Abstract: Anomaly detection on road traffic is an important task due to its great potential in urban traffic management and road safety. It is also a very challenging task since the abnormal event happens very rarely and exhibits different behaviors. In this work, we present a model to detect anomaly in road traffic by learning from the vehicle motion patterns in two distinctive yet correlated modes, i.e., the static mode and the dynamic mode, of the vehicles. The static mode analysis of the vehicles is learned from the background modeling followed by vehicle detection procedure to find the abnormal vehicles that keep still on the road. The dynamic mode analysis of the vehicles is learned from detected and tracked vehicle trajectories to find the abnormal trajectory which is aberrant from the dominant motion patterns. The results from the dual-mode analyses are finally fused together by driven a re-identification model to obtain the final anomaly. Experimental results on the Track 2 testing set of NVIDIA AI CITY CHALLENGE show the effectiveness of the proposed dual-mode learning model and its robustness in different real scenes. Our result ranks the first place on the final Leaderboard of the Track 2.

Vehicle control device, vehicle control method, and vehicle control program

Abstract: A driver performs switching to a manual driving mode in a state that the driver is suitable for a driving operation In a vehicle that is driven by a plurality of driving modes including a self-driving mode to be performed by a self-driving control unit and a manual driving mode in which a driver performs a part of or all of a driving operation, before the manual driving mode is started, information for presenting a request for operating by the driver from a user interface unit to the driver is output to the user interface unit The input unit receives a signal based on the operation by the driver A notification unit notifies a self-driving control unit of a switching signal for instructing switching to the manual driving mode when a difference between a value obtained from the signal based on the operation by the driver, the signal having been input from the input unit and a reference value according to the operation request is within an allowable range

Objective Micro-Facial Movement Detection Using FACS-Based Regions and Baseline Evaluation

Abstract: Micro-facial expressions are regarded as an important human behavioural event that can highlight emotional deception. Spotting these movements is difficult for humans and machines, however research into using computer vision to detect subtle facial expressions is growing in popularity. This paper proposes an individualised baseline micro-movement detection method using 3D Histogram of Oriented Gradients (3D HOG) temporal difference method. We define a face template consisting of 26 regions based on the Facial Action Coding System (FACS). We extract the temporal features of each region using 3D HOG. Then, we use Chi-square distance to find subtle facial motion in the local regions. Finally, an automatic peak detector is used to detect micro-movements above the proposed adaptive baseline threshold. The performance is validated on two FACS coded datasets: SAMM and CASME II. This objective method focuses on the movement of the 26 face regions. When comparing with the ground truth, the best result was an AUC of 0.7512 and 0.7261 on SAMM and CASME II, respectively. The results show that 3D HOG outperformed for micro-movement detection, compared to state-of-the-art feature representations: Local Binary Patterns in Three Orthogonal Planes and Histograms of Oriented Optical Flow.

Largest highly efficient 203 × 203 mm2 CH3NH3PbI3 perovskite solar modules

Abstract: We have been working on the fabrication of practical perovskite solar modules using an existing thin-film solar module patterning technology. Patterning processes were applied using laser scribing and mechanical scribing on a glass/F-doped SnO2 (FTO)/compact TiO2 (c-TiO2)/meso-TiO2/CH3NH3PbI3/2,2',7,7'-tetrakis(N,N-di-p-methoxyphenylamine)-9,9'-spirobifluorene/Au structure. However, TiO2 remained in the FTO–Au junction at the cell connection part: this appeared to create a series resistance that degraded the performance of the module. UV laser treatment was examined for application to the cell connection part to minimize this resistance. The results showed that the resistivity of the cell connection part, which was conventionally 0.26 Ωcm2, was reduced to 0.1 Ωcm2 by this laser treatment. We designed the solar module based on the value of a contact resistivity of 0.1 Ωcm2 and the current–voltage curve of a small single cell. As a result of prototyping solar cell modules (35 cells in series) with an aperture area of 354 cm2 on a 203 × 203 mm2 glass substrate, a module efficiency of 12.6% was achieved. There appears to be no serious impediment to the fabrication of the perovskite solar modules.

Distributional patterns of item responses and total scores on the PHQ-9 in the general population: data from the National Health and Nutrition Examination Survey.

Abstract: Recently, item responses and total scores on depression screening scales have been reported to have characteristic distributions in the general population. The distributional pattern of responses to the Patient Health Questionnaire-9 (PHQ-9) in the general population has not been well studied. Thus, we carried out a pattern analysis of the PHQ-9 item responses and total scores in US adults. Data (5372 individuals) were drawn from the 2013–2014 National Health and Nutrition Examination Survey in the United States. The item responses and total score distributions of the PHQ-9 data were investigated with graphical analysis and exponential regression model. Lines of item responses showed the same pattern among the nine items, characterized by crossing at a single point between “not at all” and “several days” and a parallel pattern from “several days” to “nearly every day” on a log-normal scale. The total score distribution of the PHQ-9 exhibited an exponential pattern, except for at the lower end of the distribution. The present results support that the item responses and total scores on the PHQ-9 in the general population show the same characteristic patterns, consistent with the previous studies using the Center for Epidemiologic Studies Depression Scale (CES-D) and Kessler Screening Scale for Psychological Distress (K6).

High power 3-phase to 3-phase matrix converter using dual-gate GaN bidirectional switches

Abstract: Highly efficient three-phase to three-phase matrix converters using Gallium nitride (GaN) bidirectional switches with both high current and high breakdown voltage are demonstrated. The GaN switch with dual gates works as a bidirectional switch by a single device, while a conventional bidirectional switch consists of four devices by two Insulated Gate Bipolar Transistors (IGBTs) and two diodes. In addition, the GaN bidirectional switch is also free from the voltage offsets for the current conduction so that the GaN-based matrix converter enables small size and highly efficient AC/AC conversion. Improvement of the device performance including the introduction of the recessed gate enables the low on-state resistance with stable operation free from current collapse. The maximum drain current reaches 100 A together with the breakdown voltage of 1340 V. The fabricated three-phase to three-phase matrix converter exhibits the maximum conversion efficiency of 98% at 1 kW output power with the expectation that the maximum output power can reach 10 kW or more by the high current device.

Nonaqueous electrolyte battery

Abstract: This nonaqueous electrolyte battery comprises a positive electrode, a negative electrode, a separator that is interposed between the positive electrode and the negative electrode, and a nonaqueous electrolyte. The positive electrode contains vanadium pentoxide, a first carbon black and a first binder; and the negative electrode contains a silicon-containing material, graphite, a second carbon black and a second binder. When the battery voltage is at 0 V, the potentials of the positive electrode and the negative electrode are respectively at 2.5-3.5 V with respect to lithium metal.

Stability of the Distribution of Patient Health Questionnaire-9 Scores Against Age in the General Population: Data From the National Health and Nutrition Examination Survey.

Abstract: Background: Epidemiological studies using the nine-item Patient Health Questionnaire (PHQ-9) have reported inconsistencies regarding the relationship between age and total scores. To determine whether this discrepancy is due to the stability of the distribution of PHQ-9 total scores against age, we investigated whether the total score distribution remains stable during adulthood, and also investigated the mathematical patterns of the total score distribution. Methods: The present study utilized data from 15,847 participants of the 2009-2014 United States National Health and Nutrition Examination Survey, all of whom responded to all PHQ-9 items. The stability of the total score distribution among different age groups was examined using overlap coefficients and graphical analysis. Results: High overlap coefficients were observed between all age groups for the distributions of PHQ-9 total scores, suggesting that the distribution of PHQ-9 total scores remains stable against age. Graphical analysis demonstrated that distributions of PHQ-9 total scores were similar across age groups. In addition, distributions of PHQ-9 total scores exhibited an exponential pattern, except at the lower end of the distribution. Conclusions: Our findings indicate that the stability of the distribution of PHQ-9 total scores throughout adulthood may underlie inconsistencies in the evidence regarding age-related changes in total depression scores.