Domain adaptation enables the learner to safely generalize into novel environments by mitigating domain shifts across distributions. Previous works may not effectively uncover underlying reasons that would lead drastic model degradation on target task. In this paper, we empirically reveal erratic discrimination of mainly stems from its much smaller feature norms with respect source domain. To end, propose a parameter-free Adaptive Feature Norm approach. We demonstrate progressively adapting...

This paper proposes a novel deep architecture to address multi-label image recognition, fundamental and practical task towards general visual understanding. Current solutions for this usually rely on an extra step of extracting hypothesis regions (i.e., region proposals), resulting in redundant computation sub-optimal performance. In work, we achieve the interpretable contextualized classification by developing recurrent memorized-attention module. module consists two alternately performed...

Current neural networks for 3D object recognition are vulnerable to rotation. Existing works mostly rely on massive amounts of rotation-augmented data alleviate the problem, which lacks solid guarantee rotation invariance. In this paper, we address issue by introducing a novel point cloud representation that can be mathematically proved rigorously rotation-invariant, i.e., identical clouds in different orientations unified as unique and consistent representation. Moreover, proposed is...

The integration of micro-electro-mechanical system (MEMS) with metamaterial has provided a novel route to achieve programmability via its reconfigurable capabilities. Here, we propose and demonstrate MEMS-based metadevice by using switchable winding-shaped cantilever (WCM) for active logical modulation. WCM can be actuated external driving voltage, the modulation bit is performed releasing MEMS cantilevers represent “on” “off” states. While underneath substrate surface rough after...

We focus on Unsupervised Domain Adaptation (UDA) for the task of semantic segmentation. Recently, adversarial alignment has been widely adopted to match marginal distribution feature representations across two domains globally. However, this strategy fails in adapting tail classes or small objects segmentation since objective is dominated by head categories large objects. In contrast alignment, we propose explicitly train a domain-invariant classifier generating and defensing against...

In this work, an actively tunable optical metadevice is proposed to realize multifunctional application by integrating self-assembly electrothermal actuator (ETA) and magnetic metamaterial. The released frame of ETA can control the deformed height metamaterial plate then provide high robustness. By driving external electromagnetic field, actuated rotated a force manipulate incident response. transmission intensities be gradually increased characterize "on" "off" states. Along with inputs...

Vanadium dioxide (VO2) based metamaterial perfect absorbers (MPAs) have high potential application values in sensing gas molecules. However, a tuning mechanism via temperature manipulation lacks the compatibility with electronic devices. In this study, voltage-controllable device is proposed by integrating an MPA and micro-electro-mechanical system (MEMS) microheater for CO2 application. The composed of metal-dielectric-metal (MDM) structure tailored to form H-shaped metamaterial. central...

In recent years, tunable metamaterials have attracted intensive research interest due to their outstanding characteristics, which are dependent on the geometrical dimensions rather than material composition of nanostructure. Among tuning approaches, micro-electro-mechanical systems (MEMS) is a well-known technology that mechanically reconfigures metamaterial unit cells. this study, development MEMS-based reviewed and analyzed based several types actuators, including electrothermal,...

An ultra-narrowband terahertz (THz) perfect absorber with metal-insulator-metal (PA-MIM) microstructures is presented. MIM composed of Al disk on SiO2/Al/Si substrate. This configuration will induce dipole-image interaction and cause an electromagnetic confinement between to eliminate the reflection. Therefore, design can absorb most THz wave realize PA. The absorption intensity proposed device could be reached 99% ultra-high Q-factor 143. By modifying gap top bottom disks MIM, response...

We present an on-chip tunable infrared (IR) metamaterial emitter for gas sensing applications. The proposed exhibits high electrical-thermal-optical efficiency, which can be realized by the integration of microelectromechanical system (MEMS) microheaters and IR metamaterials. According to blackbody radiation law, high-efficiency generated driving a Direct Current (DC) bias voltage on microheater. MEMS microheater has Peano-shaped microstructure, great heating uniformity energy conversion...

We propose and demonstrate a metamaterial design by integrating microelectromechanical system (MEMS) electrothermal actuator (ETA) platform cross-shaped (CSM) to perform opto-logic function characteristics. Reconfigurable stretchable mechanisms of CSM are achieved driving different DC bias voltages on ETA improve the limitations induced conventional use flexible substrate. The optical responses tunable electrical signals inputs. By voltage 0.20 V, tuning range is 0.54 THz obtained it...

In-situ passivation of Fe nanoparticles exsolved from perovskite cathodes for improving catalytic activity CO 2 electrolysis.

The integration of a high-performance metamaterial (MM) onto mechanically flexible and deformable substrates offers significant promise in electronics. Here, we propose two types stretchable infrared (IR) MMs to design tunable perfect absorber with ring-shape (PA-RS) cross-shape (PA-CS) on PDMS/Au/PDMS substrate, respectively. By stretching devices along different directions, PA-RS PA-CS exhibit ultra-narrowband, polarization-dependent/independent, switchable characterizations the IR...

We propose four types of metamaterial absorbers (MAs) to have tunable absorption resonance and polarization-dependent/independent, single-band, dual-band switch characterizations. By tailoring MAs with different heights between top bottom metal layers, can be modified at a 5.92 μm wavelength, the intensity attenuated gradually. These characteristics MA devices potentially used in variable optical attenuator (VOA). changing distance unit cells devices, resonances tuned range 1.74 wavelength...

Abstract A tunable terahertz (THz) chain-link metamaterial (CLM) is presented, which composed of a tailored Au layer fabricated on Si substrate. CLM exhibits bidirectional polarization-dependent characteristic by applying direct-current (dc) bias voltage device. This device can be heated up the surrounding temperature to tune corresponding resonance. The tuning range 0.027 THz from 0.318 0.291 0.60 V 1.32 V. By reconfiguring gap between CLM, there are single-resonance with red-shift at TE...

A design of electrostatically tunable metamaterial is proposed in the terahertz (THz) frequency range using electric split-ring resonators (eSRR), which shows polarization-dependent and switching characteristics. The eSRR composed anchored rectangular a movable middle I-shaped metal bar. By exploiting electrostatic force to change gap within distance bar, electromagnetic characteristics can be controlled. response single-resonance transverse (TE) mode keeps as quite stable. In magnetic (TM)...