基于改进MobileFaceNet的人脸识别方法北大核心

针对目前大多数人脸识别算法参数多、计算量大,难以部署到移动端和嵌入式设备中的问题,提出了一种基于改进MobileFaceNet的人脸识别方法。通过对MobileFaceNet模型结构的调整,将bottleneck模块优化为sandglass模块,改良深度卷积和逐点卷积的相对位置,适当增大sandglass模块的输出通道数,从而减少特征压缩时的信息丢失,增强人脸空间特征的提取。实验结果表明:改进后的方法在LFW测试数据集上准确率达99.15%,模型大小和计算量分别仅为原算法的61%和45%,验证了所提方法的有效性。     

Mutual information maximizing GAN inversion for real face with identity preservation

Recent generative adversarial networks (GANs) have yielded remarkable performance in face image synthesis. GAN inversion embeds an image into the latent space of a pretrained generator, enabling it to be used for real face manipulation. However, current inversion approaches for real faces suffer the dilemma of initialization collapse and identity loss. In this paper, we propose a hierarchical GAN inversion for real faces with identity preservation based on mutual information maximization. We first use a facial domain guaranteed initialization to avoid the initialization collapse. Furthermore, we prove that maximizing the mutual information between inverted faces and their identities is equivalent to minimizing the distance between identity features from inverted and original faces. Optimization for real face inversion with identity preservation is implemented on this mutual information-maximizing constraint. Extensive experimental results show that our approach outperforms state-of-the-art solutions for inverting and editing real faces, particularly in terms of face identity preservation.     

Structural regulation and polishing performance of dendritic mesoporous silica (D-mSiO2) supported with samarium-doped cerium oxide composites

Ceria (CeO₂) particles are prevalent polishing abrasive materials. Trivalent lanthanide ions are the popular category of dopants for enriched surface defects and thus improved physicochemical properties, since they are highly compatible with CeO₂ lattices. Herein, a series of dendritic-like mesoporous silica (D-mSiO₂)-supported samarium (Sm)-doped CeO₂ nanocrystals were synthesized via a facile chemical precipitation method. The relation of the structural characteristics and chemical mechanical polishing (CMP) performances were investigated to explore the effect of Sm-doping amounts on the D-mSiO₂/Sm_xCe_1?xO_2?δ (x = 0–1) composite abrasives. The involved low-modulus D-mSiO₂ cores aimed to eliminate surface scratch and damage, resulting from the optimized contact behavior between abrasives and surfaces. The trivalent cerium (Ce³⁺) and oxygen vacancy (V_O) at CeO₂ surfaces were expected to be reactive sites for the material removal process over SiO₂ films. The optimal oxide-CMP performances in terms of removal efficiency and surface quality were achieved by the 40% Sm-doped composite abrasives. It might be attributed to the high Ce³⁺ and V_O concentrations and the enhancement of tribochemical reactivity between CeO₂SiO₂ interfaces. Furthermore, the relationship between the surface chemistry, polishing performance as well as the actual role in oxide-CMP of the D-mSiO₂/Sm_xCe_1?xO_2?δ abrasives were also discussed.     

Dielectric performance and conductive mechanism of KNN-based ceramics prepared via sol–gel core–shell process

Journal of Materials Science: Materials in Electronics - 0.95(Li0.02Na0.50K0.48)(Nb0.95Sb0.05)O3–0.05AgTaO3@BaZrO3 (LNKNbSAT@BZ) lead-free ceramics were prepared via a sol–gel...     

Real time power management strategy for fuel cell hybrid electric bus based on Lyapunov stability theorem

The fuel cell/battery durability and hybrid system stability are major considerations for the power management of fuel cell hybrid electric bus (FCHEB) operating on complicated driving conditions. In this paper, a real time nonlinear adaptive control (NAC) with stability analyze is formulated for power management of FCHEB. Firstly, the mathematical model of hybrid power system is analyzed, which is established for control-oriented design. Furthermore, the NAC-based strategy with quadratic Lyapunov function is set up to guarantee the stability of closed-loop power system, and the power split between fuel cell and battery is controlled with the durability consideration. Finally, two real-time power management strategies, state machine control (SMC) and fuzzy logic control (FLC), are implemented to evaluate the performance of NAC-based strategy, and the simulation results suggest that the guaranteed stability of NAC-based strategy can efficiently prolong fuel cell/battery lifespan and provide better fuel consumption economy for FCHEB.     

Dynamic simulation of hydrogen-based off-grid zero energy buildings with hydrogen storage considering Fanger model thermal comfort

In this study, some locations with different climates, off-grid zero energy buildings with hydrogen energy storage systems are designed, and transient analysis is conducted. These considered buildings supply their electricity consumption without using the electrical grid and PV panels or wind turbines. Also, they supply thermal comfort to occupants by using a vapor compression chiller and humidifier. Domestic hot water of occupants is supplied using solar collectors. For analyzing building's performance and objectives achievement, TRNSYS software is used. Also, for evaluating occupant thermal comfort, the Fanger model is used. The considered building is a one-story building with a 150 m² area. Four occupants are considered. Both of them are seated at rest, and another is seated with light working such as typing. Using the Fanger model equation and MATLAB software, the thermal comfort of occupants is determined. For domestic hot water consumption, verified profiles that vary during 24 h of the day are considered. Achieved results show that for humid and cold cities, PV panels with an area of 73 and 76 m² can be supplied the required electricity of considered building with four occupants and battery state of charge is higher than 50% and 10%, respectively. Moreover, with a suitable air conditioner system, the predicted percentage of dissatisfied (PPD) can be lower than 12% and 8% for humid and cold cities. Therefore, the building can be converted to a zero-energy building using its rooftop area.     

Fabrication and properties of diatomite ceramics with hierarchical pores based on direct stereolithography

Porous diatomite ceramics with hierarchical pores and high apparent porosity (50.29–56%) were successfully fabricated via direct stereolithography. The pre-ball-milling time, dispersant type and dispersant concentration were systematically investigated to prepare diatomite pastes with high solid loading, low viscosity and a self-supporting effect. The results showed that a pre-ball-milling time of 24 h was more suitable to prepare diatomite pastes with high solid loading, and Span80 at 2 wt% was the optimal dispersant to obtain 40 vol% diatomite paste with a low viscosity and a self-supporting effect. To restrain the formation of defects, a heating rate as low as 0.2 °C/min was allowed to control the pyrolysis rate in the multistage debinding process. At sintering temperatures ranging from 900 °C to 1000 °C, porous diatomite ceramics exhibited a typical bimodal porosity, high apparent porosity and great flexural strength.     

Effect of ligand environment of rare-earth ions on temperature measurement performance of SrAO4:xEu3+ (A = Mo and W) phosphors

Molybdate and tungstate with scheelite-type structure are excellent self-luminescent materials, which can be used as ideal hosts for the doping of rare-earth ions. In this study, a series of Eu³⁺-activated SrAO₄ (A = Mo and W) phosphors were successfully synthesized, and their crystal structures, photoluminescence properties, and temperature measurement performance were analyzed in detail. These phosphors were excited by UV light (291 nm and 247 nm, respectively), with clear energy transfer (ET) (MoO₄^2?→Eu³⁺ or WO₄^2?→Eu³⁺). According to fluorescence intensity ratio (FIR) and Judd–Ofelt (J–O) theory, compared to SrWO₄:0.01Eu³⁺ phosphor, SrMoO₄:0.01Eu³⁺ phosphor exhibited better thermal stability, with relatively low Sa value (maximum values were 5.082 %K^?1 and 20.74 %K^?1, respectively), and their Sr values were not significantly different (maximum values were 0.864 %K^?1 and 0.83 %K^?1, respectively). Sa value was negatively correlated to central asymmetry of Eu³⁺, but the optimal Sr value tended to be more suitable for central asymmetry of Eu³⁺. In addition, Eu³⁺ exhibited stronger central asymmetry as well as covalency of Eu–O bond in SrMoO₄. Results reveal that SrMoO₄:xEu³⁺ and SrWO₄:xEu³⁺ can be used for luminescent thermometers.     

High-entropy titanate pyrochlore as newly low-thermal conductivity ceramics

Low-thermal conductivity ceramics play an indispensable role in maximizing the efficiency and durability of hot end components. Pyrochlore, particularly zirconate pyrochlore, is currently a highly promising and widely studied candidate for its extremely low thermal conductivity. However, there are still few pyrochlores that offer both stiffness, insulation, and good thermal expansion properties. In this work, the solidification method was innovatively introduced into the preparation of titanate pyrochlore, and combined it with the compositional design of high-entropy. Through careful composition design and solidification control, the high-density and uniform elements distributed high-entropy titanate pyrochlore ceramics were successfully prepared. These samples possess high hardness (15.88 GPa) and Young’s modulus (295.5 GPa), low thermal conductivity (0.947 W·m^?1·K^?1), excellent thermal expansion coefficient (11.6 ×10^?6/K) and an exquisite balance between stiffness and insulation (E/κ, 312.1 GPa·W^?1·m·K), in which the E/κ exhibits the highest value among the current reported works.