Search

Single domain generalization (SDG) is a more realistic setting than domain generalization (DG) and domain adaptation (DA). It aims to train a domain-agnostic model in the presence of a single source domain to perform well on arbitrary unseen target domains. To facilitate the practical deployment of remote sensing image segmentation in the real world, we propose a novel SDG method, termed category consistency on domain randomization (CCDR). To expand the coverage of a single source domain, CCDR implements a simple yet effective data generation module to perform domain randomization of texture and style information, since texture divergences from different geographical environments or phenological periods, and style divergences from different illumination or weather conditions, are the main causes of the domain shift in remote sensing. In addition to emphasizing inter and intraclass relationships within each domain via multidomain supervised learning, CCDR further draws inspiration from the triple loss to enhance the semantic correlation across the source domain and the generated auxiliary domain, which makes the segmentation model more sensitive to class discriminative information and better adapted to unseen target domains. In comparison to other state-of-the-art SDG methods, CCDR can synthesize the more effective auxiliary domain, and perform more reliable classification on the unseen target domains without the sophisticated training pipeline and cumbersome data generation process. The experimental results on two public remote sensing datasets demonstrate that CCDR has remarkable advantages in remote sensing image segmentation tasks. And the code is publicly available at: https://github.com/LCB1970/CCDR.

Selective hydrogenation is a significantly important catalytic process in many industrial processes, where Pd-based catalysts are the most commonly used. Herein, the surface properties of γ-Al2O3were modified by the layered flower-like architectures of alkaline MgO using a simple hydrothermal method. Then, a series of Pd/Al2O3–MgO catalysts were prepared by an incipient wetness impregnation method for the selective hydrogenation of isoprene. The results showed that the Pd/Al2O3–0.12MgO catalyst exhibited superior catalytic performance with a high activity (93%) and high selectivity (99%) for the semi-hydrogenation products. Systematic studies have demonstrated that the support effect plays an important role in dispersing and stabilizing palladium particles. The reasons for the improved catalytic performance were attributed to the high dispersion of Pd nanoparticles, the changed electronic structure of Pd, and the strong interaction between Pd nanoparticles and supports. This work opens a simple route to regulate the electronic properties of active metal sites viasupport modification for selective hydrogenation reactions.

This paper proposes an advanced approach for detecting faces with mask occlusion based on YOLOv5 to address various challenges encountered in face detection, including illumination blur and occlusion. The proposed methodology involves the integration of a convolutional block attention module into the backbone network and different network levels in the neck of the YOLOv5s. This approach enables the suppression of irrelevant features and emphasizes the identification of masked facial features. Replacing the conventional loss function with the Focal Loss function addresses the problem of sample imbalance. The enhanced YOLOv5s network was applied to detect mask-occluded faces. Empirical evaluations were conducted on the WIDER Face and AIZOO datasets. The simulation comparison results demonstrate that the proposed method achieves superior real-time detection performance, fulfilling the objective of developing a lightweight detection model.

Despite achieving exceptional performance, deep neural networks (DNNs) suffer from the harassment caused by adversarial examples, which are produced by corrupting clean examples with tiny perturbations. Many powerful defense methods have been presented such as training data augmentation and input reconstruction which, however, usually rely on the prior knowledge of the targeted models or attacks. In this paper, we propose a novel approach for detecting adversarial images, which can protect any pre-trained DNN classifiers and resist an endless stream of new attacks. Specifically, we first adopt a dual autoencoder to project images to a latent space. The dual autoencoder uses the self-supervised learning to ensure that small modifications to samples do not significantly alter their latent representations. Next, the mutual information neural estimation is utilized to enhance the discrimination of the latent representations. We then leverage the prior distribution matching to regularize the latent representations. To easily compare the representations of examples in the two spaces, and not rely on the prior knowledge of the targeted model, a simple fully connected neural network is used to embed the learned representations into an eigenspace, which is consistent with the output eigenspace of the targeted model. Through the distribution similarity of an input example in the two eigenspaces, we can judge whether the input example is adversarial or not. Extensive experiments on MNIST, CIFAR-10, and ImageNet show that the proposed method has superior defense performance and transferability than state-of-the-arts.

Due to more abundant data sources, more various objects of interest, and more time-consuming annotations, there is a large amount of out-of-distribution (OOD) data in the remote sensing field, on which the performance of high-accuracy image segmentation models trained under ideal experimental conditions generally degrades dramatically. Domain adaptation (DA) consequently comes into being, which aims to learn the predictor for the label-scarce target domain of interest with the help of the label-sufficient source domain in the presence of the distribution difference, namely, domain shift, between the two domains. However, the off-the-shelf DA methods for image segmentation not only struggle to cope with the more complex domain shift problems in remote sensing imagery but also almost cannot process heterogeneous data directly without information loss. While the current heterogeneous DA methods mostly still rely on some supervision information from the target domain, which is typically inaccessible in the real world. To overcome these drawbacks, we propose the multilevel heterogeneous unsupervised DA (UDA) method, termed MHDA, which unifies the instance-level DA based on cycle consistency, the feature-level DA based on contrastive learning, and the decision-level DA based on task consistency into a framework to more effectively handle the complex domain shift and heterogeneous data. After that, extensive DA experiments are conducted on the International Society for Photogrammetry and Remote Sensing (ISPRS) dataset, the BigCity dataset constructed by ourselves, and the Wuhan University (WHU) dataset, to explore the effect of each module in MHDA, the necessity of heterogeneous DA, and the effectiveness of multilevel DA. And the results demonstrate that MHDA can achieve superior performance on the remote sensing image segmentation task, compared with several state-of-the-art DA methods.

Due to insufficient information and feature extraction in existing face-detection methods, as well as limited computing power, designing high-precision and efficient face-detection algorithms is an open challenge. Based on this, we propose an improved face detection algorithm. First, through 1 × 1’s common convolution block (CBL) expands the channel for feature extraction, introduces a depthwise separable residual network into the YOLO-v4 network to further reduce the amount of model computation, and uses CBL to reduce the dimension, so as to improve the efficiency of the subsequent network. Second, the improved attention mechanism is used to splice the high-level features, and the high-level features and the shallow features are fused to obtain the feature vectors containing more information, so as to improve the richness and representativeness of the feature vectors. Finally, the experimental results show that compared with other comparative methods, our method achieves the best results on public face datasets, and our performance in personal face detection is significantly better than other methods.

The efficient electrochemical conversion of carbon dioxide (CO2) to carbon monoxide (CO) using renewable energy is an effective route to pursue carbon neutrality. Optimizing the binding energy of CO on palladium (Pd) metal-based materials used in this process is to make sure the timely desorption of CO from their active sites is critical. Tuning the electronic structure of the Pd center is an effective strategy to optimize its catalytic performance. Herein, we rationally design Pd nanoparticles (NPs)/polymeric carbon nitride (CN) (Pd/CN) composite, which alters the electronic structure of Pd by introducing the interfacial polarization effect to accelerate CO desorption and improve CO selectivity of Pd catalyst. The optimized Pd/CN exhibits a CO Faradaic efficiency of 92.7% at −0.9 V versusreversible hydrogen electrode in CO2-saturated 0.1 M KHCO3solution. Experimental investigations and theoretical calculations jointly confirm that the enhanced CO selectivity and stability originate from the electron transfer at the Pd/CN interface, and the weakened *CO adsorption on the palladium hydride surface.

ABSTRACTLung adenocarcinoma (LUAD) represents the major histological type of lung cancer with high mortality globally. Due to the heterogeneous nature, the same treatment strategy to various patients may result in different therapeutic responses. Hence, we aimed to elaborate an effective signature for predicting patient survival outcomes. The TCGA-LUAD cohort from the TCGA portal was used as a training dataset. The GSE26939 and GSE68465 cohorts from the GEO database were taken as validation datasets. All immunologically relevant genes were extracted from the ImmPort. The ESTIMATE algorithm was employed to explore LUAD microenvironment in the training dataset. Further, the DEGs were picked out based on the immune-associated genes reflecting different statuses in the immune context of TME. Univariate/multivariate Cox regression was performed to determine six prognosis- specific genes (PIK3CG, BTK, VEGFD, INHA, INSL4, and PTPRC) and established a risk predictive signature. The time-dependent ROC indicated that AUC values were all greater than 0.70 at 1-, 3-, and 5- year intervals. Corresponding RiskScore of each LUAD patient was calculated from the signature, and they were stratified into the high- and low-risk groups by the median value of RiskScore. K-M curves and Log-rank test demonstrated significant survival differences between the two groups (P< 0.05). Similar results were exhibited in the validation datasets. The RiskScore was incredibly relevant to clinicopathological factors like gender, AJCC stage, and T stage. Also, it can mirror the distribution state of 15 kinds of TIICs and have some predictive value for the sensitivity of therapeutic drugs.

The low redox potential of conjugated donor–acceptor (D–A) type polymer due to strong electron interaction between donor units and acceptor units is benefit to increase the power density, which is suitable to serve as the anode materials of lithium-ion battery. In this paper, the D–A type polymers containing phenanthraquinone (PQ) as acceptor unit and thiophene as donor unit was used as anode materials of lithium-ion battery. Three polymers with different content of donor unit are synthesized to study the effects of donor unit content on the electrochemical properties. It is found that the low content of donor unit in the polymer is benefit to the specific capacities, but not benefit to the rate capability of the electrode. The polymers with the donor–acceptor ratio of 1:1, 2:1 and 3:1 display discharge specific capacity of 737 mAh•g−1, 552 mAh•g−1and 499 mAh•g−1at the current density of 72 mA•g−1, respectively. The rate capability of polymer with higher donor–acceptor ratio is better due to its lower band gap and better conductivity. The three D–A type molecules all have perfect cycle stability and stable molecular structure. For a situation of C=C group in aromatic rings partly stored lithium ions, one simple method is proposed for the first time to confirm the exactly lithium ion storage sites by using the Natural Bond Orbital (NBO) analysis. It is found that the lithium ions are mainly stored in the PQ units (redox sites A, RSA) and the thiophene units adhered to the PQ unit directly (redox sites B, RSB), while the thiophene units not adhere to the PQ units did not storage Li-ion.

In this study, unique mesoporous structure alumina microspheres, each of which comprises an assembly of numerous highly crystallized alumina nanorods, were prepared viaa facile template-free hydrothermal method. Such alumina microspheres were further assembled in one step to form a macroporous structure with abundant inter-microspheres viapore-directing-agent-free molding process. Owing to the morphological merits and the remarkable mechanical stability of the as-formed alumina microspheres, the macropores formed from the stable packing of the alumina microspheres had average sizes as large as 265 nm and were well connected with each other. Moreover, the inter-microsphere mesopores with an average size of 12.2 nm were well-maintained after the molding process. With such hierarchical porosity and large specific surface area, the as-prepared alumina support enabled better dispersion of the active MoNi components. Accordingly, the macro–mesoporous MoNi/Al2O3catalyst exhibited both significantly higher catalytic activities and longer service life for the hydrodemetallization of nickel(ii) 5, 10, 15, 20-tetraphenylporphine and for the hydrodesulfurization of dibenzothiophene as compared to controls with a single type of pores. With the ease of preparation, large surface area, and remarkable activity and stability, such a macro–mesoporous alumina support provides a promising alternative for hydrotreating heavy raw oils in industry.

New approach to activating carbon active sites to facilitate the electroconversion of CO2through N-vacancy engineering is proposed and confirmed by the density functional theory calculations and experimental results. N-vacancy engineered graphitic C3N4(g-C3N4) is identified as an efficient electrocatalyst to boost CH4formation owing to the three-coordinating to two-coordinating transition of the carbon atoms that surround N vacancies. The defected g-C3N4(DCN) exhibits a 44% faradaic efficiency with a CH4partial current density of 14.8 mA/cm2at −1.27 VRHEin 0.5 M KHCO3electrolyte, exceeding all the reported carbon-based materials and even being comparative to Cu-based electrocatalysts. Creation of more unsaturated carbon atoms enables the harmonic energy overlap near band gap edge between DCN and *CO, accounting for an improved strength of *CO on DCN, a lower energy barrier and an enhanced CO2RR to form CH4. This strategy holds the promise for tuning atomic configuration to enhance activity of catalyst.