Search

Small object detection is a fundamental and challenging issue in computer vision. We believe that there are two factors that affect the performance of small object detection: small object dataset and small object itself. In terms of datasets, we introduce a dataset named SeaDefine, which opens up a new direction for small object detection in maritime environment. For the small object itself, we utilize deep feature learning and feature fusion network (DFLFFN) to help detect objects. Concretely, the designed deep feature learning module (DFLM) at the single-layer level can describe objects for a variety of scenarios through activating multi-scale receptive fields over a wider scope. Meanwhile, to intensify classification capacity of small objects, the shallow features with rich details will be integrated with the deep features generated from the DFLM by introducing feature fusion block (FFB). In addition, we analyze the multi-scale strategy from a mathematical perspective to a certain extent. A large number of results in the experiments show that proposed DFLFFN achieves the leading detection performance on the MS-COCO and SeaDefine datasets. In particular, our DFLFFN surpasses the baseline by 4.1 points on APrs score for SeaDefine dataset, and 7.8 points on AP S score for MS-COCO dataset. [ABSTRACT FROM AUTHOR]

Small object detection is a challenging task in computer vision. We claim that the huge performance gap between the small object detectors and normal sized object detectors stems from two aspects, including the small object dataset and the small object itself. In terms of datasets, we build a large-scale dataset with high image resolution dubbed Small-PCB, in order to promote detection in semiconductor industry. For the small object itself, we utilize multi-scale feature learning and feature fusion strategy to help detect objects. More concretely, we devise two novel components to predict small objects better: re-parameterized module with channel shuffle (RMCS) and multi-scale feature enhanced convolution (MFEC). MFEC aims to split input channels into several parts and applies convolutions with different sizes to each part, and adopt point-by-point convolution to fuse individual channel features. RMCS not only use structural re-parameterization, but also channel shuffle. The usage of channel shuffle can be seen as a fusion of channel features. It strengthens feature information interaction between different channel groups, which bring more informative feature clues. Based on the RMCS and the MFEC, we introduce OIU-RMCS and M-MFEC, respectively. Finally, we build our I-YOLO via integrating these two components into a YOLO-based detector. A large number of qualitative and quantitative results in the experiments indicate that our proposed I-YOLO achieves the state-of-the-art performance on the popular AI-TODv2 and Small-PCB datasets. [ABSTRACT FROM AUTHOR]

• Rethinking the PASCAL-VOC and MS-COCO dataset for small object detection. • There are different labeling errors in these two benchmarks by visual analysis. • Building several specific datasets to analyze small object detection. • Data quality is another factor that affects the performance of small object detection. The data and the algorithm are critical to deep learning-based small object detectors. In this paper, we rethink the PASCAL-VOC and MS-COCO dataset for small object detection. By visual analysis of the original annotations, we find that there are different labeling errors in these two datasets. To solve these problems, we build specific datasets, including SDOD, Mini6K, Mini2022 and Mini6KClean. The experimental results of several typical algorithms (e.g. SSD, YOLOv5, Faster RCNN and Deformable DETR) on the datasets show that data labeling errors (such as missing labels, category label errors, inappropriate labels) are another factor that affects the detection performance of small objects. [ABSTRACT FROM AUTHOR]

Increasing demand for electricity has placed heavy stress on power system security. Therefore, this paper focuses on the problem of how to maximize power system static security in terms of branch loading and voltage level under normal operation and even the most critical single line contingency conditions. This paper proposes a hybrid approach to find out the optimal locations and settings of two classical types of flexible AC transmission system (FACTS) devices, namely thyristor-controlled series compensators (TCSCs) and static var compensators (SVCs) for solving this problem. Our proposed approach requires a two-step strategy. Firstly, the min cut algorithm (MCA) and tangent vector technique (TVT) are applied to determine the proper candidate locations of TCSC and SVC respectively so as to reduce the search scope for a solution to the problem, and then the cuckoo search algorithm (CSA) is employed to solve this problem by simultaneously optimizing the locations and settings for TCSC and SVC installation. The proposed hybrid approach has been verified on the IEEE 6-bus and modified IEEE 14-bus test systems. The results indicate that CSA outperforms particle swarm optimization (PSO), proving its effectiveness and potential, and they also show that our proposed hybrid approach can find the best locations and settings for TCSC and SVC devices as an effective way for enhancing power system static security by removing or alleviating the overloads and voltage violations under normal operation and even the most critical single line contingency conditions. Using this hybrid approach, the search space for solution to the problem becomes limited hence the computational burden will be decreased. [ABSTRACT FROM AUTHOR]

Due to the limitations of LiDAR, such as its high cost, short service life and massive volume, visual sensors with their lightweight and low cost are attracting more and more attention and becoming a research hotspot. As the hardware computation power and deep learning develop by leaps and bounds, new methods and ideas for dealing with visual simultaneous localization and mapping (VSLAM) problems have emerged. This paper systematically reviews the VSLAM methods based on deep learning. We briefly review the development process of VSLAM and introduce its fundamental principles and framework. Then, we focus on the integration of deep learning and VSLAM from three aspects: visual odometry (VO), loop closure detection, and mapping. We summarize and analyze the contribution and weakness of each algorithm in detail. In addition, we also provide a summary of widely used datasets and evaluation metrics. Finally, we discuss the open problems and future directions of combining VSLAM with deep learning. [ABSTRACT FROM AUTHOR]

This paper is devoted to the study of a fully discrete T-ψ finite element method based on H-decomposition to solve a nonlinear degenerate transient eddy current problem with ferromagnetic materials. Here, the ferromagnetic properties are linked by a power material law. We first design a nonlinear time-discrete scheme for approximation in suitable function spaces. We show the well-posedness of the semidiscrete problem and prove the convergence of the nonlinear scheme by the Minty-Browder technique. Finally, we suggest a fully discrete scheme, derive its error estimate and give some numerical experiments to validate the theoretical result. [ABSTRACT FROM AUTHOR]

Detecting small or tiny objects is always a difficult and challenging issue in computer vision. In this paper, we provide a latest and comprehensive survey of deep learning-based detection approaches from the perspective of small or tiny objects. Our survey is featured by thorough and exhaustive analysis of small or tiny object detection. We comprehensively introduce 30 existing datasets about small or tiny objects, and summarize different definitions of small or tiny objects based on different application scenarios, such as pedestrian detection, traffic signs detection, face detection, remote sensing target detection and object detection in common life. Then small or tiny object detection techniques are overviewed systematically from seven aspects, including super-resolution techniques, context-based information, multi-scale representation learning, anchor mechanism, training strategy, data augmentation, and schemes based on loss function. Finally, the detection performance of small or tiny objects on 12 popular datasets is analyzed in depth. Based on performance analysis, we also discuss the promising research directions in the future. We hope this survey could provide researchers guidance to catalyze understanding of small or tiny object detection and further facilitate research on small or tiny object detection systems. • Summarize the 30 datasets about small or tiny objects. • Provide definitions of small/tiny objects based on different application scenarios. • Systematically review small/tiny object detection techniques from seven aspects. • Present the evaluation results of different methods for small/tiny object detection. • Discuss future research directions of small/tiny object detection. [ABSTRACT FROM AUTHOR]

Defect detection in complex background is a critical issue. To address this issue, this paper proposes the mixture attention mechanism cascade network, in which the new channel attention network is linked with the spatial attention network to create an effective mixed attention network that takes advantage of their respective advantages, adaptively suppresses background noise, and highlights defect features. To ensure the efficiency and effectiveness of effective mixed attention network, the new channel attention network splices the output features of the global average pooling layer and the global maximum pooling layer and then sends the spliced features into a shared network, which is a one‐dimensional convolutional network, and uses cross‐channel interaction for fusion. Furthermore, in order to provide more discriminative feature representation, the authors extract the intermediate features of the region proposal network and input them into effective mixed attention network. Finally, the cascade head is used to refine the predicted bounding box to achieve high‐quality defect location. To demonstrate the superiority and usefulness of this method, it is compared to the latest method using widely used PCB and NEU data sets. A large number of trials demonstrate that this strategy outperforms other methods for detecting defects in complicated backgrounds. [ABSTRACT FROM AUTHOR]

Fully discrete potential-based finite element methods called $${\mathbf{A}-\phi}$$ methods are used to solve a transient eddy current problem in a three-dimensional convex bounded polyhedron. Using $${\mathbf{A}-\phi}$$ methods, fully discrete coupled and decoupled numerical schemes are developed. The existence and uniqueness of solutions for these schemes together with the energy-norm error estimates are provided. To verify the validity of both schemes, some computer simulations are performed for the model from TEAM Workshop Problem 7. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the finite element A -- ø method to approximate the eddy current equations with discontinuous coefficients in general three-dimensional Lipschitz polyhedral eddy current region. Nonmatching finite element meshes on the interface are considered and optimal error estimates are obtained. [ABSTRACT FROM AUTHOR]

A posteriori error estimate of the discontinuous-streamline diffusion method for first-order hyperbolic equations was presented, which can be used to adjust space mesh reasonably. A numerical example is given to illustrate the accuracy and feasibility of this method. [ABSTRACT FROM AUTHOR]

Small object detection is a challenging problem in computer vision. It has been widely applied in defense military, transportation, industry, etc. To facilitate in-depth understanding of small object detection, we comprehensively review the existing small object detection methods based on deep learning from five aspects, including multi-scale feature learning, data augmentation, training strategy, context-based detection and GAN-based detection. Then, we thoroughly analyze the performance of some typical small object detection algorithms on popular datasets, such as MS-COCO, PASCAL-VOC. Finally, the possible research directions in the future are pointed out from five perspectives: emerging small object detection datasets and benchmarks, multi-task joint learning and optimization, information transmission, weakly supervised small object detection methods and framework for small object detection task. [ABSTRACT FROM AUTHOR]

[Display omitted] • Synthesis of soluble TCRA to insoluble HCP-TCRA by Friedel-Crafts alkylation reaction. • HCP-TCRA has strong hydrophobicity, which is conducive to the enrichment of SES in aqueous solutions. • The SPME fibers were used repeatedly for more than 30 times without any decrease in extraction effect. • The SPME fiber could extract estrogen directly from simple pretreated milk samples. A novel hyper-crosslinked 2, 8, 14, 20-tetraphenyl calix [4] resorcinol arene polymer (HCP-TCRA) was synthesized by using an external crosslinking of 2, 8, 14, 20-Tetraphenyl calix [4] resorcinol arene (TCRA) as monomers and dimethoxymethane as crosslinker. The HCP-TCRA was explored as a SPME fiber coating material. A variety of physicochemical characterization demonstrated that the material has excellent chemical/thermal stability and strong hydrophobicity. The fiber demonstrated fantastic adsorptive property for three synthetic estrogens (diethylstilbestrol, hexestrol, ethinyloestradiol, abbreviated to SESs). In this work, an efficient HCP-TCRA fiber-based SPME method was developed to analyze SESs by coupling with HPLC-MS detection. Under the optimized SPME conditions, this method was shown a wide range of linearity (2.0–100.0 ng mL−1 for diethylstilbestrol, 1.0–100.0 ng mL−1 for hexestrol, and 30.0–480.0 ng mL−1 for ethinyloestradiol), low limits of detection (LODs, 0.5 ng mL−1 for DES 0.3 ng mL−1 for HEX, 10.0 ng mL−1 for EES) and fine reusability (more than 30 times). Finally, the proposed methods had been applied to the determination of SESs in milk samples with a simple pretreatment. At three different levels, relative recoveries presented between 86.7 and 106.4%. The method put up highly sensitive and budget-friendly in detection SESs, and has capacity in quality control of milk products. [ABSTRACT FROM AUTHOR]

The coconut hispine beetle, Brontispa longissima Gestro, is an invasive and economically important pest of palm trees in China. Beauveria bassiana has been used widely to control agriculture and forestry pests, for example, Dendrolimus punctatus Walker, Pantana phyllostachysae Chao, and Basilepta melanopus Lefevre. Until now, there is no published data on the use of B. bassiana against B. longissima. In this study, we identified highly pathogenic B. bassiana strains against different developmental stages of B. longissima under laboratory conditions. The BB-01 strain presented the highest virulence against B. longissima. After inoculating with the BB-01 strain at (1.0 ± 0.5) × 107 spores·mL−1for 10 d (28 ± 1°C with 95 ± 1% relative humidity), the corrected mortality of B. longissima adults reached 94.7% with a LT50 of 5.3 days. Mortality on other developmental stages was 84.4 ± 3.3% (LT50 = 5.8 d) for 1st-2nd instars, 76.1 ± 1.2% (LT50 = 6.71 d) for 3rd-4th instars, 73.7 ± 1.1% (LT50 = 6.98 d) for ≥5 instars, respectively. BB-01 median lethal concentrations (LC50) against 1st-2nd instar, 3rd-4th instar, ≥5 instar larvae and adults for 6 d were 9.0 × 106, 3.5 × 107, 4.8 × 107, 9.1 × 105 spores·mL−1, respectively. The BB-01 strain exhibited a lethal effect peak on B. longissima at 6–7 d after treatment. This study provides basic data about the pathogenicity of different B. bassiana strains on B. longissima. [ABSTRACT FROM AUTHOR]

• Fe, N-doped graphene wrapped on nanoparticles of carbon black. • Hierarchically microporous structure was constructed. • Graphitic N sites with more ORR activity were preferentially formed. • The catalyst exhibits an excellent ORR activity in alkaline electrolyte. Fe, N-doped graphene-like carbon layers-wrapped carbon black composite (L-Fe-AD+C) was developed by pyrolyzing the mixture of Fe-adenine precursors and conductive carbon black as an efficient catalyst to oxygen reduction reaction (ORR). Graphitic N sites with more ORR activity are preferentially formed due to the facilitation of metallic Fe. Furthermore, graphene-like ultra-thin networks wrapped around nanoparticles of carbon black greatly increase the exposure of active sites and construct continuous diffusion channels favorable for the transfer of substances. As a result, the as-prepared L-Fe-AD+C exhibits an excellent ORR activity in alkaline electrolyte with a high half wave potential (0.88 V), and superior selectivity together with admirable stability, which are comparable to those of the benchmark Pt/C catalyst. More encouragingly, the L-Fe-AD+C can serve as an efficient cathode catalyst in a zinc-air battery, which reaches a peak power density of 221 mW cm−2 approximating to that of Pt/C catalyst (236 mW cm−2). The battery also shows a steady discharge voltage plateau of 1.26 V and specific capacity of 799 mAh g−1 at a discharge current density of 10 mA cm−2. The results suggest the considerable potential of the L-Fe-AD+C catalyst to replace precious metal catalysts. [ABSTRACT FROM AUTHOR]

• Nitrogen atoms were in situ doped into carbon cloth. • The catalyst is flexible and self-standing. • The catalyst sheet can be directly as assembled in a zinc-air battery as ORR catalyst and current collector of cathode. • The catalyst is a good substrate for powder ORR catalyst. Nitrogen doped carbon materials have attracted tremendous attention as the promising candidates for platinum-based catalysts toward oxygen reduction reaction (ORR). Herein, we report a facile preparation of N-doped carbon cloth (Co-N-OCC) as a flexible and self-standing electrocatalyst toward ORR by calcining oxidized carbon cloth coated with cobalt ions in ammonia atmosphere. The pyridinic N, Co-N and pyrrolic N, related to ORR activity, are the dominant configurations of nitrogen with an atomic ratio of 42.27%, 25.36% and 20.29% in the catalyst. The resulting Co-N-OCC shows favorable ORR activity with an onset potential of 0.87 V vs. RHE, an electron transfer number of 3.57, and good methanol tolerance in alkaline electrolyte. Co-N-OCC catalyst sheets were directly assembled in zinc-air battery as catalyst and current collector of cathode. The ORR catalytic performance of Pt/C is further boosted in a zinc-air battery by using the Co-N-OCC sheet as an electrode substrate. [ABSTRACT FROM AUTHOR]

Lymphocytes express voltage-gated (Kv) and Ca2+-activated (IKCa1) K+ channels. Recently, we found that WEHI-231, an immature B cell line, expresses voltage-independent K+ channels called large-conductance background K + channels (LKbg). Arachidonic acid (AA) has attracted attention because of its potential regulatory roles in the apoptosis of immature B cells. To elucidate the functional targets of AA, we investigated the effects of AA on membrane currents, voltages, and cytoplasmic Ca2+ concentration ([Ca2+]c) of WEHI-231 and Bal-17 cells that represent immature and mature mouse B cells, respectively. In whole-cell patch clamp, both Kv and IKCa1 were inhibited by AA. On the other hand, AA activated LKbg current and non-selective cationic (NSC) current in WEHI-231 while only NSC current in Bal-17. Inside–out patch clamp study showed that AA directly activates LKbg. AA induced hyperpolarization of WEHI-231 and depolarization of Bal-17 cells, respectively. The selective functional expression of LKbg and their activation by AA were also confirmed in the immature B cells (B220+/AA4.1+) freshly isolated from mouse spleen. In fura-2 spectrofluorimetry, AA induced persistent increase in [Ca2+]c of WEHI-231 cells, which was attenuated by KCl-induced depolarization. In Bal-17 cell, however, AA induced only a transient increase of [Ca2+]c. In summary, the novel type of background K+ channels (LKbg) in immature B cells is strongly activated while the other K+ channels (Kv and IKCa1) commonly expressed in lymphocytes are inhibited by AA. The hyperpolarization and augmentation of Ca2+ influx by LKbg activation might play a role in the response of immature B cells to AA. [ABSTRACT FROM AUTHOR]