•A novel ReSidual U-block (RSU) is designed to capture multi-scale deep features.•A nested U-structure, called U2-Net, that uses RSU is developed for salient object detection.•Both large (176.3 MB) ...and small (4.7 MB) instances of U2-Net get competitive results.
In this paper, we design a simple yet powerful deep network architecture, U2-Net, for salient object detection (SOD). The architecture of our U2-Net is a two-level nested U-structure. The design has the following advantages: (1) it is able to capture more contextual information from different scales thanks to the mixture of receptive fields of different sizes in our proposed ReSidual U-blocks (RSU), (2) it increases the depth of the whole architecture without significantly increasing the computational cost because of the pooling operations used in these RSU blocks. This architecture enables us to train a deep network from scratch without using backbones from image classification tasks. We instantiate two models of the proposed architecture, U2-Net (176.3 MB, 30 FPS on GTX 1080Ti GPU) and U2-Net† (4.7 MB, 40 FPS), to facilitate the usage in different environments. Both models achieve competitive performance on six SOD datasets. The code is available: https://github.com/NathanUA/U-2-Net.
Nanocarbon materials, such as carbon nanotube and graphene, are reckoned as the ideal reinforcement for composites due to their extraordinary intrinsic properties. However, the commonly-used ...homogenous architecture could not guarantee an effective enhancement in the mechanical and physical properties and become the major technological barrier for practical applications. Fabrication of nanocarbon reinforced composites with special architecture such as laminate, alignment and network architecture has been reported to break through the limitations and improve the overall performance of the composites. In this review, we aim to: (1) systematically summarize the current preparation methods to achieve different types of architecture design in three classes of matrices (polymer, ceramic and metal); (2) analyze the mechanisms and the influence factors for the mechanical and physical properties (electrical and thermal conductivities) of the composites with different architecture types. Finally, we outline the main advances and outlooks for architecture design in nanocarbon reinforced composites.
Display omitted
Short-wavelength infrared light with a wavelength range of 1–3 µm is crucial in various fields related to human activities. It has a better penetrating ability than visible light in ...harsh weather conditions (such as nighttime, fog, and dust) and can produce higher resolution natural images compared to mid-wavelength infrared and long-wavelength infrared light. In addition, the corresponding devices can work at room temperature without refrigeration, which is beneficial to reducing the cost and the size of the equipment, since no thermal imaging is involved. To date, many reviews have discussed the progress of infrared photodetectors from the perspective of a single type of material (such as two-dimensional metal chalcogenides, colloidal quantum dots, conjugated polymers, etc.) or a single synthesis/enhancement strategy (solution-processable, surface plasmon, optical field manipulation, etc.). As discussions from a single perspective are relatively one-sided, a more comprehensive overview is required. Herein, we review the recent research progress on SWIR photodetectors, especially the advantages and disadvantages of devices based on different materials, and provide some perspectives for the future application of SWIR photodetectors.
The development of cloud infrastructures inspires the emergence of cloud-native computing. As the most promising architecture for deploying microservices, serverless computing has recently attracted ...more and more attention in both industry and academia. Due to its inherent scalability and flexibility, serverless computing becomes attractive and more pervasive for ever-growing Internet services. Despite the momentum in the cloud-native community, the existing challenges and compromises still wait for more advanced research and solutions to further explore the potential of the serverless computing model. As a contribution to this knowledge, this article surveys and elaborates the research domains in the serverless context by decoupling the architecture into four stack layers: Virtualization, Encapsule, System Orchestration, and System Coordination. Inspired by the security model, we highlight the key implications and limitations of these works in each layer, and make suggestions for potential challenges to the field of future serverless computing.
Completely Automated CNN Architecture Design Based on Blocks Sun, Yanan; Xue, Bing; Zhang, Mengjie ...
IEEE transaction on neural networks and learning systems,
2020-April, 2020-Apr, 2020-4-00, 20200401, Volume:
31, Issue:
4
Journal Article
Open access
The performance of convolutional neural networks (CNNs) highly relies on their architectures. In order to design a CNN with promising performance, extensive expertise in both CNNs and the ...investigated problem domain is required, which is not necessarily available to every interested user. To address this problem, we propose to automatically evolve CNN architectures by using a genetic algorithm (GA) based on ResNet and DenseNet blocks. The proposed algorithm is completely automatic in designing CNN architectures. In particular, neither preprocessing before it starts nor postprocessing in terms of CNNs is needed. Furthermore, the proposed algorithm does not require users with domain knowledge on CNNs, the investigated problem, or even GAs. The proposed algorithm is evaluated on the CIFAR10 and CIFAR100 benchmark data sets against 18 state-of-the-art peer competitors. Experimental results show that the proposed algorithm outperforms the state-of-the-art CNNs hand-crafted and the CNNs designed by automatic peer competitors in terms of the classification performance and achieves a competitive classification accuracy against semiautomatic peer competitors. In addition, the proposed algorithm consumes much less computational resource than most peer competitors in finding the best CNN architectures.
While cloud computing has brought paradigm shifts to computing services, researchers and developers have also found some problems inherent to its nature such as bandwidth bottleneck, communication ...overhead, and location blindness. The concept of fog/edge computing is therefore coined to extend the services from the core in cloud data centers to the edge of the network. In recent years, many systems are proposed to better serve ubiquitous smart devices closer to the user. This article provides a complete and up-to-date review of edge-oriented computing systems by encapsulating relevant proposals on their architecture features, management approaches, and design objectives.
Display omitted
Advances in miniaturized portable electronics and progress on novel enabling technologies, consequently accompanied by power consumption downgraded from the scale of milliwatts (mW) ...to microwatts (μW), have inevitably facilitate the development of an emerging discipline-wearable human energy conversion systems. Served as a passive human energy harvester which can directly convert heat into electricity in long-term operations without the user’s intervention, wearable thermoelectric generators (WTEG) have sparked considerable research interest for next-generation power supply. In comparison to the longstanding research history of thermoelectrics, their wearables are still in infancy of extensive growth over the last decade. Although, historically, the main challenge behind the conventional thermoelectric generator (TEG) is the improvement of dimensionless figure-of-merit (zT), wearable applications usually impose additional restrictions that can be more pivotal than zT value. Diversified targeted strategies therefore have been proposed to push TEG toward wearable application. Here, we review the evolutionary roadmap of the wearable thermoelectric generators in the past decade, it could be concluded that the trend in WTEG is to move toward stretchable three-dimension (3D)-structure with rational thermal design at the moment. The basic concept targeting WTEG, which highly differs from that of the traditional TEG, is introduced at first. And then, aiming to provide detailed design guidelines for WTEG, we begin with carefully discussing the key issues for TEG toward wearable application. Finally, the specific strategies targeted WTEG that is classified into thermal design regarding extrinsic temperature difference (ΔText), parasitic and TEG thermal resistance, mechanical design with emphasis on optimizing deformability at materials/device level beyond flexibility toward stretchability, as well as architecture design from two-dimension (2D) to 3D feature are comprehensively summarized, respectively. With these understandings, perspectives for the future development of WTEG are outlined. This review emphasizes issues and provides additional insight in advanced strategies for pushing TEG toward wearable application. The key issues clarified and the design roadmap summarized here arise from the goal of providing ideas for the concurrent optimization of the future WTEG, as well as realistically promoting the TEG toward wearable application.