RESUMEN

Enhancing the robustness of vision algorithms in real-world scenarios is challenging. One reason is that existing robustness benchmarks are limited, as they either rely on synthetic data or ignore the effects of individual nuisance factors. We introduce OOD-CV-v2, a benchmark dataset that includes out-of-distribution examples of 10 object categories in terms of pose, shape, texture, context and the weather conditions, and enables benchmarking of models for image classification, object detection, and 3D pose estimation. In addition to this novel dataset, we contribute extensive experiments using popular baseline methods, which reveal that: 1) Some nuisance factors have a much stronger negative effect on the performance compared to others, also depending on the vision task. 2) Current approaches to enhance robustness have only marginal effects, and can even reduce robustness. 3) We do not observe significant differences between convolutional and transformer architectures. We believe our dataset provides a rich test bed to study robustness and will help push forward research in this area. Our dataset is publically available online, https://genintel.mpi-inf.mpg.de/ood-cv-v2.html.

RESUMEN

The scaling of bulk Si-based transistors has reached its limits, while novel architectures such as FinFETs and GAAFETs face challenges in sub-10 nm nodes due to complex fabrication processes and severe drain-induced barrier lowering (DIBL) effects. An effective strategy to avoid short-channel effects (SCEs) is the integration of low-dimensional materials into novel device architectures, leveraging the coupling between multiple gates to achieve efficient electrostatic control of the channel. We employed TCAD simulations to model multi-gate FETs based on various dimensional systems and comprehensively investigated electric fields, potentials, current densities, and electron densities within the devices. Through continuous parameter scaling and extracting the sub-threshold swing (SS) and DIBL from the electrical outputs, we offered optimal MoS2 layer numbers and single-walled carbon nanotube (SWCNT) diameters, as well as designed structures for multi-gate FETs based on monolayer MoS2, identifying dual-gate transistors as suitable for high-speed switching applications. Comparing the switching performance of two device types at the same node revealed CNT's advantages as a channel material in mitigating SCEs at sub-3 nm nodes. We validated the performance enhancement of 2D materials in the novel device architecture and reduced the complexity of the related experimental processes. Consequently, our research provides crucial insights for designing next-generation high-performance transistors based on low-dimensional materials at the scaling limit.

RESUMEN

We employ a combination of accelerated molecular dynamics and machine learning to unravel how the dynamic characteristics of CBL-B and C-CBL confer their binding affinity and selectivity for ligands from subtle structural disparities within their binding pockets and dissociation pathways. Our predictive model of dissociation rate constants (k off) demonstrates a moderate correlation between predicted k off and experimental IC50 values, which is consistent with experimental k off and τ-random accelerated molecular dynamics (τRAMD) results. By employing a linear regression of dissociation trajectories, we identified key amino acids in binding pockets and along the dissociation paths responsible for activity and selectivity. These amino acids are statistically significant in achieving activity and selectivity and contribute to the primary structural discrepancies between CBL-B and C-CBL. Moreover, the binding free energies calculated from molecular mechanics with generalized Born and surface area solvation (MM/GBSA) highlight the ΔG difference between CBL-B and C-CBL. The k off prediction, together with the key amino acids, provides important guides for designing drugs with high selectivity.

RESUMEN

The human-machine interface (HMI) previously relied on a single perception interface that cannot realize three-dimensional (3D) interaction and convenient and accurate interaction in multiple scenes. Here, we propose a collaborative interface including electrooculography (EOG) and tactile perception for fast and accurate 3D human-machine interaction. The EOG signals are mainly used for fast, convenient, and contactless 2D (XY-axis) interaction, and the tactile sensing interface is mainly utilized for complex 2D movement control and Z-axis control in the 3D interaction. The honeycomb graphene electrodes for the EOG signal acquisition and tactile sensing array are prepared by a laser-induced process. Two pairs of ultrathin and breathable honeycomb graphene electrodes are attached around the eyes for monitoring nine different eye movements. A machine learning algorithm is designed to train and classify the nine different eye movements with an average prediction accuracy of 92.6%. Furthermore, an ultrathin (90 µm), stretchable (∼1000%), and flexible tactile sensing interface assembled by a pair of 4 × 4 planar electrode arrays is attached to the arm for 2D movement control and Z-axis interaction, which can realize single-point, multipoint and sliding touch functions. Consequently, the tactile sensing interface can achieve eight directions control and even more complex movement trajectory control. Meanwhile, the flexible and ultrathin tactile sensor exhibits an ultrahigh sensitivity of 1.428 kPa-1 in the pressure range 0-300 Pa with long-term response stability and repeatability. Therefore, the collaboration between EOG and the tactile perception interface will play an important role in rapid and accurate 3D human-machine interaction.

Asunto(s)

RESUMEN

High-performance electromagnetic interference (EMI) shielding materials with ultralow density, excellent flexibility, and good mechanical properties are highly desirable for aerospace and wearable electronics. Herein, honeycomb porous graphene (HPG) fabricated by laser scribing technology is reported for EMI shielding and wearable applications. Due to the honeycomb structure, the HPG exhibits an EMI shielding effectiveness (SE) up to 45 dB at a thickness of 48.3 µm. The single-piece HPG exhibits an ultrahigh absolute shielding effectiveness (SSE/t) of 240 123 dB cm2/g with an ultralow density of 0.0388 g/cm3, which is significantly superior to the reported materials such as carbon-based, MXene, and metal materials. Furthermore, MXene and AgNWs are employed to cover the honeycomb holes of the HPG to enhance surface reflection; thus, the SSE/t of the HPG/AgNWs composite membrane can reach up to 292 754 dB cm2/g. More importantly, the HPG exhibits excellent mechanical stability and durability in cyclic stretching and bending, which can be used to monitor weak physiological signals such as pulse, respiration, and laryngeal movement of humans. Therefore, the lightweight and flexible HPG exhibits excellent EMI shielding performance and mechanical properties, along with its low cost and ease of mass production, which is promising for practical applications in EMI shielding and wearable electronics.

Asunto(s)

RESUMEN

Effective battery thermal management (BTM) is critical to ensure fast charging/discharging, safe, and efficient operation of batteries by regulating their working temperatures within an optimal range. However, the existing BTM methods not only are limited by a large space, weight, and energy consumption but also hardly overcome the contradiction of battery cooling at high temperatures and battery heating at low temperatures. Here we propose a near-zero-energy smart battery thermal management (SBTM) strategy for both passive heating and cooling based on sorption energy harvesting from air. The sorption-induced reversible thermal effects due to metal-organic framework water vapor desorption/sorption automatically enable battery cooling and heating depending on the local battery temperature. We demonstrate that a self-adaptive SBTM device with MIL-101(Cr)@carbon foam can control the battery temperature below 45 °C, even at high charge/discharge rates in hot environments, and realize self-preheating to ∼15 °C in cold environments, with an increase in the battery capacity of 9.2%. Our approach offers a promising route to achieving compact, liquid-free, high-energy/power-density, low-energy consumption, and self-adaptive smart thermal management for thermo-related devices.