Segregation is a key challenge in promoting more diverse and inclusive cities. Research based on large-scale mobility data indicates that segregation between majority and minority groups persists in daily activities b...
详细信息
The problem of detecting not occupied areas of the road is useful in many ADAS functions that include Lane Departure Warning (LDW), Lateral Control (LC), Adaptive Cruise Control (ACC). The problem may be in general de...
The problem of detecting not occupied areas of the road is useful in many ADAS functions that include Lane Departure Warning (LDW), Lateral Control (LC), Adaptive Cruise Control (ACC). The problem may be in general defined as the ability to distinguish such areas from those that are occupied by other traffic members. The objective here is to provide data useful, for example, in planning vehicle's maneuvers. In this work, we propose a solution that supports the recognition of such areas. The proposed method is based on a procedure similar to the one used in Discrete Wavelet Transform (DWT) used in image decomposition. One of the introduced modifications relies on using simpler filters, whose structure facilitates their hardware implementation.
The dynamic changes of task requirement and the time-varying distribution of resources in Industrial Internet of Things (IIoT) make a challenge for traditional static resource allocation methods to flexibly adapt to t...
The dynamic changes of task requirement and the time-varying distribution of resources in Industrial Internet of Things (IIoT) make a challenge for traditional static resource allocation methods to flexibly adapt to these changes. These can lead to increase latency and energy consumption, which result in low efficiency of resource allocation. In this paper, a resource allocation method based on multi-agent reinforcement learning (MARL) in end-edge-cloud enabled IIoT is proposed. The method builds an end-edge-cloud collaboration resource management model in IIoT scenarios, and constructs the optimization problem of minimizing task latency and energy consumption. Then, the optimization problem is further transformed into a multi-agent Markov decision process (MDP). Furthermore, the multi-agent deep deterministic policy gradient (MADDPG) is adopted to solve the formulated MDP problem. Finally, simulation results demonstrate that the proposed algorithm can significantly reduce the task latency cost by 27% and energy consumption cost by 15% compared with that of the baseline methods.
In this work, we propose a solution in which vehicle-to-infrastructure (V2I) technology could be used to provide virtual regular tracks, which would allow the vehicle to be kept on a specific lane with a specific accu...
In this work, we propose a solution in which vehicle-to-infrastructure (V2I) technology could be used to provide virtual regular tracks, which would allow the vehicle to be kept on a specific lane with a specific accuracy for a longer stretch of road or in specific points/areas, e.g. in cities, at intersections or roundabouts. The proposed methods may in the future support selected functions belonging to advanced driver assistance system (ADAS), the operation of which requires precise determination of the position of the vehicle on the road. In this work, we focus on solutions that can be cheap to produce, assemble the elements of the system, and then maintain it in operation. One of the key aspects is to achieve small sizes and low energy consumption of the wireless sensors which will form the basis of its operation. The solutions proposed in this work can be treated as theoretical considerations, as so far there is no global V2I system in which they could be fully verified.
The purpose of this paper is to examine and contrast the Novel Circular Patch Antenna employing FR4 and RT Droid (CPA). Resonant frequency of 2.5 GHz in a static medium. To create the antenna, we use the HFSS and two ...
详细信息
Exploring the expected quantizing scheme with suitable mixed-precision policy is the key point to compress deep neural networks (DNNs) in high efficiency and accuracy. This exploration implies heavy workloads for doma...
详细信息
Exploring the expected quantizing scheme with suitable mixed-precision policy is the key point to compress deep neural networks (DNNs) in high efficiency and accuracy. This exploration implies heavy workloads for domain experts, and an automatic compression method is needed. However, the huge search space of the automatic method introduces plenty of computing budgets that make the automatic process challenging to be applied in real scenarios. In this paper, we propose an end-to-end framework named AutoQNN, for automatically quantizing different layers utilizing different schemes and bitwidths without any human labor. AutoQNN can seek desirable quantizing schemes and mixed-precision policies for mainstream DNN models efficiently by involving three techniques: quantizing scheme search (QSS), quantizing precision learning (QPL), and quantized architecture generation (QAG). QSS introduces five quantizing schemes and defines three new schemes as a candidate set for scheme search, and then uses the differentiable neural architecture search (DNAS) algorithm to seek the layer- or model-desired scheme from the set. QPL is the first method to learn mixed-precision policies by reparameterizing the bitwidths of quantizing schemes, to the best of our knowledge. QPL optimizes both classification loss and precision loss of DNNs efficiently and obtains the relatively optimal mixed-precision model within limited model size and memory footprint. QAG is designed to convert arbitrary architectures into corresponding quantized ones without manual intervention, to facilitate end-to-end neural network quantization. We have implemented AutoQNN and integrated it into Keras. Extensive experiments demonstrate that AutoQNN can consistently outperform state-of-the-art quantization. For 2-bit weight and activation of AlexNet and ResNet18, AutoQNN can achieve the accuracy results of 59.75% and 68.86%, respectively, and obtain accuracy improvements by up to 1.65% and 1.74%, respectively, compared
While memristive devices are highly attractive as memory cells, they are also capable of performing computations, paving the way to futuristic in-memory computing architecture. Several memristive logic families have b...
While memristive devices are highly attractive as memory cells, they are also capable of performing computations, paving the way to futuristic in-memory computing architecture. Several memristive logic families have been proposed, and approaches to map gate-level logic circuits to such memristive implementations have been introduced. In this paper, we focus on the CRS logic family that offers several advantages compared to the more often considered IMPLY and MAGIC families. A central feature of CRS is the ability of one physical memristive device to realize varying logic gates over several clock cycles. Our method computes a schedule, i.e., an assignment which logic gate of a given circuit is executed on which memristive device during which clock cycles. Using an optimal MaxSAT model, it can minimize the resulting schedule’s duration (depth), the cost of the used memristors, or the cost of additional cache register cells, while satisfying all dependencies needed for correct computation. In addition to results of the scheduling procedure itself, we report a physical experiment that demonstrates one of the schedules and discuss the energy benefits of the CRS family.
Users' viewing behavior could affect their perception and evaluation of design works. Taking into account users' visual attention as a subjective cognition cue, we used eye-tracking evidence to identify users&...
Users' viewing behavior could affect their perception and evaluation of design works. Taking into account users' visual attention as a subjective cognition cue, we used eye-tracking evidence to identify users' focus areas for further analysis. We conducted experiments to extract the image features of design images and the reviewers' eye-tracking data, aiming to predict the product design ranking in the competition through fusion data analysis. In particular, we collected 1,504 product design images from a design competition. Four deep convolutional neural networks were selected to explore the best aesthetics computation model. The experimental results show that using design images and eye-tracking data fusion can improve the model prediction performance. Finally, MobileNet-V3 achieves the highest classification accuracy of 74.75%. This suggests the proposed method can provide useful insights into personalized aesthetics evaluation and user-centered design perception.
In ophthalmology, early fundus screening is an economic and effective way to prevent blindness caused by ophthalmic diseases. Clinically, due to the lack of medical resources, manual diagnosis is time-consuming and ma...
详细信息
Socio-spatial segregation is the physical separation of different social, economic, or demographic groups within a geographic space, often resulting in unequal access to resources, services, and opportunities. The lit...
详细信息
暂无评论