In this paper, we propose a method that addresses two challenges faced by existing generalized few-shot object detection methods, limiting their scalability in practical applications: (i) the high training cost required for few-shot learning of a small number of target object samples, and (ii) the lack of variation in these samples. Leveraging the generalization ability of a deep neural network (DNN) trained with vision-language pretraining, we introduce a frozen network few-shot object detection framework in which a common DNN, shared with the pretraining phase, is used to model the appearance features of target objects without needing to be updated during few-shot learning. In experiments, we employed publicly available datasets with diverse object appearances for pretraining and few-shot learning. By comparing our proposed method’s detection accuracy with existing methods, we demonstrate that our method effectively addresses the challenges described above.

In situations where caregivers cannot always accompany visually impaired individuals, a system that can automatically detect and convey the position of objects is necessary. In previous research on navigation systems for the visually impaired, special equipment was required, and the accuracy was insufficient to provide timely notification. Additionally, systems that notify users of the position of objects on a desk have been proposed, using markers attached to hands or dishes, which are recognized by a camera. However, position detection using markers involves the hassle of marker installation and relies on special sensors, making it difficult for everyday use. In this study, we propose a method that detects the position of objects on a screen using deep learning-based object detection, and then converts the position into spatial coordinates using projection transformation, conveying the location information to visually impaired individuals via a clock position system. The proposed method is implemented as a smartphone application, making it easy to install without the need for special equipment. This system has two applications: alerting users to the presence of obstacles in front of them when walking outdoors, and informing users of the positions of dishes on a table during meals. In experiments measuring the positions of objects in front of the user during walking, people and bicycles up to 15 m away could be detected, with a positioning error rate of 11%. In evaluation experiments measuring the positions of objects on a desk, the accuracy of positioning was up to an error of 1.8 cm, and the conversion to clock positions was 98% accurate.

Traffic warning systems have attracted attention for reducing traffic accidents caused by motorcycles. The authors previously developed a highly portable system that uses an in-vehicle web camera and a smartphone to detect motorcycles approaching at high speed based on rear-camera input images and notifies the driver of danger. In this study, we propose a system that combines deep-learning-based object detection and heuristics to detect encroaching vehicles from images taken by a front-facing in-vehicle camera. First, we consider the shape of the bounding box of the detected vehicle to identify the encroaching vehicle. Since the encroaching vehicle is turned sideways when entering the main road, we identify side-facing vehicles by calculating the aspect ratio of the bounding box. For motorcycles, the criterion for distinguishing between parked motorcycles and encroaching motorcycles is the presence or absence of a rider. To distinguish between parked vehicles and moving vehicles, the values of the background pixels around the detected vehicle are compared with the values of pixels around the corresponding vehicle in the previous frame and the difference is used to determine whether the vehicle is moving. The results show that the above heuristics improve the accuracy of detecting encroaching vehicles.

To reduce motorcycle accidents, recent efforts have focused on traffic warning systems. In this study, we developed a highly portable system that utilizes smartphones and web cameras to detect motorcycles approaching at high speeds and to notify drivers of potential dangers. Specifically, we have implemented a feature to detect hazardous scenarios where motorcycles change lanes from behind and overtake on the right. Accurate lane estimation is crucial for precisely assessing situations where motorcycles change lanes. While lane estimation methods based on recognizing white road markings exist, they may not be effective on roads with unclear or faded white lines, which are common in developing countries. An alternative approach involves estimating lanes from fixed camera vehicle trajectories, but this is limited to locations where fixed cameras are installed. Moreover, deep learning methods for lane detection face computational challenges on mobile systems, especially when analyzing both lane positions and vehicle behavior. Here we propose a method to detect vehicles and estimate lanes based on their trajectories from videos captured by in-vehicle cameras in developing countries. The effectiveness of our proposed method was demonstrated through real-world road tests in Indonesia.

This paper simultaneously addresses three limitations associated with conventional skeleton-based action recognition; skeleton detection and tracking errors, poor variety of the targeted actions, as well as person-wise and framewise action recognition. A point cloud deep-learning paradigm is introduced to the action recognition, and a unified framework along with a novel deep neural network architecture called Structured Keypoint Pooling is proposed. The proposed method sparsely aggregates keypoint features in a cascaded manner based on prior knowledge of the data structure (which is inherent in skeletons), such as the instances and frames to which each keypoint belongs, and achieves robustness against input errors. Its less constrained and tracking-free architecture enables time-series keypoints consisting of human skeletons and nonhuman object contours to be efficiently treated as an input 3D point cloud and extends the variety of the targeted action. Furthermore, we propose a Pooling-Switching Trick inspired by Structured Keypoint Pooling. This trick switches the pooling kernels between the training and inference phases to detect person-wise and framewise actions in a weakly supervised manner using only video-level action labels. This trick enables our training scheme to naturally introduce novel data augmentation, which mixes multiple point clouds extracted from different videos. In the experiments, we comprehensively verify the effectiveness of the proposed method against the limitations, and the method outperforms state-of-the-art skeleton-based action recognition and spatio-temporal action localization methods.

This study investigates unsupervised anomaly action recognition, which identifies video-level abnormal-human-behavior events in an unsupervised manner without abnormal samples, and simultaneously addresses three limitations in the conventional skeleton-based approaches: target domain-dependent DNN training, robustness against skeleton errors, and a lack of normal samples. We present a unified, user prompt-guided zero-shot learning framework using a target domain-independent skeleton feature extractor, which is pretrained on a large-scale action recognition dataset. Particularly, during the training phase using normal samples, the method models the distribution of skeleton features of the normal actions while freezing the weights of the DNNs and estimates the anomaly score using this distribution in the inference phase. Additionally, to increase robustness against skeleton errors, we introduce a DNN architecture inspired by a point cloud deep learning paradigm, which sparsely propagates the features between Joints. Furthermore, to prevent the unobserved normal actions from being misidentified as abnormal actions, we incorporate a similarity score between the user prompt embeddings and skeleton features aligned in the common space into the anomaly score, which indirectly supplements normal actions. On two publicly available datasets, we conduct experiments to test the effectiveness of the proposed method with respect to above-mentioned limitations.

To reduce sideswipes and collision accidents involving two- and four-wheeled vehicles under mixed traffic flow conditions, we previously created a smartphone application (app) that predicts acceleration and driving lane behaviors of two-wheeled vehicles. In this system, vehicles are detected from road images taken by a smartphone camera, and vehicles positions on the road are estimated by our projection conversion algorithm. However, regarding that app, it is necessary to improve the accuracy of the vanishing point calculations in the camera images. Accordingly, in order to reduce calculation costs, we created a method that integrates road segmentation and vehicle detection to create a new scheme for detecting road edges and the vanishing point, even on roads without lane lines. These improvements will help maintain the accuracy of vanishing point calculations while facilitating their high real-time characteristics.

Since the 1980s, the single-core optical fiber cable has been the essential data transmission wiring media, with ongoing increases in deployments. Its use case has broadened, with transmission distances ranging from meters to 10000s km, including intrabuilding, fiber-to-the-home (FTTH), terrestrial, and even submarine trunk lines. Currently, optical fiber cable deployment is strongly driven by cloud-based storage, X as a Service (XaaS), and worldwide data center (DC) functionality with the optical interconnection of massive servers on DC campuses. Hence, it is quite natural that the optical fiber cable industry has developed ultrahigh fiber count, high density, and small diameter optical fiber cable, e.g., 6912-fiber-count cable. With high-density cables, massive data capacity can be installed at one time while saving limited space in an underground duct. High-density cable designs have been developed not just by cable structure optimization but through the process of bundling fibers and the use of high-grade optical fibers. Moreover, many enabling technologies have been developed to support the deployment of ultrahigh fiber count cables, such as fiber identification, connectors, mass fusion splicing, and storage systems for fusion splicing points. In this article, state-of-the-art technologies and prospects for ultrahigh-count and ultrahigh-density optical fiber cables and enabling technologies are explained.

A self-driving system for an electric wheelchair must estimate its position, identify the travelable area, and determine the travel route. Indoors, self-location estimation can be performed by wireless LAN or beacon. However, for self-driving, more accurate position estimation is required. In previous research on self-driving of electric wheelchairs, there are methods using LIDAR (Light Detection and Ranging) and laser range scanner. However, the conventional research has a problem that the system configuration is complicated and the power consumption also increases. In this study, we developed position estimation by deep learning using smartphone images with a positioning accuracy that allowed self-driving at low cost. Also, when the smartphone image recognition does not work, such as when the wheelchair approaches a wall, position estimation by augmented reality (AR) markers was used. In addition, a Bluetooth low-energy beacon was used to approximate the position when no marker could be seen. The route from an initial position to a destination was calculated using map information stored within the system. Experiments confirmed that the proposed system allows self-driving of an electric wheelchair along the route from its initial position to a destination. In this prototype system, we used the GPU board to calculate deep learning, but in the future we plan to concentrate the system on smart phones by calculating with the AI chip embedded in the smartphone.

Solar power generation has attracted significant attention recently as a safe and environmentally friendly renewable energy source. However, generally speaking, since the service lives of solar power systems are relatively long, and since it is difficult to detect anomalies in individual solar panels, such plants tend to operate without much consideration for individual panel anomalies. In order to more comprehensively monitor solar power generation systems, the National Institute of Advanced Industrial Science and Technology (AIST) of Japan has developed a direct current (DC) power line communication system that enables monitoring of each panel in a system. Monitored data are then integrated and uploaded to the cloud. Using this monitored data, we found that the integrated power ratio trends of single panel power output are in accordance with a normal distribution. Therefore, herein, we propose an anomaly detection method that uses a normal distribution. We then describe an experiment using 24 solar panels into which pseudo-faults were induced and show that our proposal makes it possible to detect errors with high accuracy.

Rechargeable sensor networks have been studied as a candidate for next generation wireless sensor networks. However, rechargeable sensor nodes cannot always communicate with other nodes because the rechargeable sensor nodes cannot receive a stable supply power. To cope with this problem, transmission power control algorithm RT-TPC (Relay Traffic-based Transmission Power Control) has been proposed. However, the algorithm does not consider the power generation pattern of each sensor node. In addition, RT-TPC is only able to support the node placement of a one-dimensional problem. In this paper, we propose a two-dimensional routing protocol based on the power generation pattern of the sensor node. Furthermore, we also propose a synchronization protocol to support the uneven charge-rate of the sensor node. In our protocol, each sensor node sends its power generation pattern and charge rate to the neighboring node. The information is used to decide the selection the destination node. The preliminary evaluation shows that the proposed protocols are able to achieve higher delivery ratio than the previous work.

The rapid growth of communication networks such as the Internet and mobile networks have spurred the development of numerous online trading systems. The important point on the trade in such a community is the formation of trust with trading partners. System that stores the information about the trust and reputation and forms the online credit is a reputation system. However, malicious evaluator can enter an incorrect evaluation value to the reputation system. Such attacks are called the undercover marketing or the stealth marketing. In order to resist the attacks of the undercover marketing, reputation systems considering the evaluation capability of evaluators have been proposed. However, the system could not sufficiently respond to attacks that perform undercover marketing in pinpoint after having increased the level of evaluation capability. This paper proposes a novel algorithm to calculate the evaluation capability which reduces the influence of the value which is out of the daily trends. Simulation results show that the proposed method reduces the influence of the undercover marketing.

Recently, highly accurate positioning devices enables us to provide various types of location based services (LBS). On the other hand, because such positioning data include deeply personal information, the protection of location privacy is one of the most significant problems in LBS. Lots of different techniques for securing the location privacy have been proposed, for instance the concept of Silent period, the concept of Dummy node, and the concept of Cloaking-region. However, many of these researches have a problem that quality of the LBS (QoS) decreased when anonymity is improved, and anonymity falls down when QoS is improved. In this paper, we present a node density-based location privacy scheme which can provide location privacy by utilizing hybrid concept of Dummy node and Cloaking-region. Simulation results show that the probability of tracking of a target node by an adversary is reduced and the QoS of LBS is also improved.

The Bloom filter is a data structure which the characteristics of contents is expressed by a bit pattern. The Bloom filter has become attractive as one of the methods of looking up data in the distributed system. Since two or more keywords can be used in the method based on the Bloom filter, it is more flexible than the distributed hash table (DHT). Especially, the key aspect of the structured Bloom filter is that the number of query forwarding is fixed. In the previous research, the ring of the Bloom filter, which is like Chord ring, was proposed. However, the node in the ring must maintain many filters corresponding to the number of nodes. In this paper, we propose the tree structured Bloom filter and evaluated the performance of the query forwarding. We also propose similarity-based tree management method which is effective to reduce the cost of reconstruction of the tree. To evaluate our method, the size of filter information and the number of average hops of query forwarding in this research are compared to the existing research. The tree reconstruction cost is also evaluated by simulation.

A Bloom filter offers a simple space-efficient randomized data structure for representing a set in order to support membership queries. The Bloom filter has become attractive as one of the methods of looking up data in the distributed system. Since two or more keywords can be used in the method based on the Bloom filter, it is more flexible than the distributed hash table (DHT). Especially, the key aspect of the structured Bloom filter is that the number of query forwarding is fixed. In the research of the past, the ring of the Bloom filter which is like the Chord structure is proposed. However, the node in the ring must maintain many filters corresponding to the number of nodes. In this paper, we propose the tree structured Bloom filter and evaluated the performance of the query forwarding. The size of filter information and the number of average hops of query forwarding in this research are compared to the existing research. The filter information managed in this research is reduced, and the number of hops is smaller than existing research.

The Bloom filter has become attractive as one of the methods of looking up data in the distributed system. Since two or more keywords can be used in the method based on the Bloom filter, it is more flexible than the distributed hash table (DHT). Especially, the key aspect of the structured Bloom filter is that the number of query forwarding is fixed. In the research of the past, the ring of the Bloom filter which is like the Chord structure is proposed. However, the node in the ring must maintain many filters corresponding to the number of nodes. In this research, the Bloom filter based on the tree structure is proposed. The filter information managed in the node is reduced, and the size of information in this research is compared to the existing research.

In the community of P2P, it is paid attention to try to control illegal behavior of the peer by managing the reputation of the peer. However, the implementation overhead was large, and the semantics of the trust model was indefinite in many of proposed methods. On the other hand, the trustworthiness estimation method by a simple, probabilistic technique, that is, by the maximum likelihood estimation is proposed. If this method is used, the problem concerning the semantics of the reliability model and the problem of the implementation overhead can be improved. However, the defense to a collusive attack is not proposed in this method. In this paper, it proposes the trust estimation method of the peer to give the defense to a collusive attack to the maximum likelihood estimation method. Moreover, peer's trustworthiness is correctly estimated though there is a collusive attack in this proposal method from the simulation result

In ad-hoc multicast networks, the maintenance of the delivery tree is important. However, the cost of reconstruction of the delivery tree is very large. In this paper, we propose the method to recover the delivery tree locally and effective retransmission for the packet loss caused by disconnection of the delivery tree. A new node group named the support group is composed in surroundings of the delivery tree for a local recovery. The support group is maintained by recursive constructing in the surroundings though the member of this support group might be included in the delivery tree by a local recovery. From our simulation, the proposed method has better performance than ODMRP.

A remote monitoring of various device distributed over wide area is strongly demanded to cut down their maintenance cost. With the introduction of wireless packet systems, the mobile remote monitoring system has started to propagate, because of a less limit to establish remote monitoring devices and its characteristic of always connectivity. Many small to medium sized corporations have been hesitating, although they understand this necessity, because of high investment costs involved in the maintenance of monitoring systems. One solution to this problem is the ASP (Application Service Provider). We developed the multi sharing mobile remote monitoring system MMRM. This paper will present the four layers object management model and the three layers application protocol stuck model that are basic design schemes of MMRM. The system structure and the function block diagram of the practical system TRM (Tokai-Remote Monitoring System) in which Toshiba Machine's TC-mini is used as a data logger will be introduced. In addition, system characteristics such as the delay period from trouble detection to sending alarm message will also be described.

In the field of mobile communication, not only has the scale of network environments increased over the last few years, but there has also been a steady increase in the diversity of mobile communication protocols and service applications due to the spread of mobile Internet applications. To enable these networks to be designed and developed efficiently, it is therefore necessary to have some way of simulating entire systems in order to evaluate different schemes, predict the traffic levels and try out alternative designs. In this study, we simulate a large-scale wide-area mobile Internet environment by allowing multiple existing simulators to cooperate with each other. Specifically, by calling a wireless data link protocol software simulator on an OPNET from an ns-2 simulator running a wide-area network simulation, we have conducted a simulated evaluation of a TCP/IP communication protocol suite including a W-CDMA environment, thereby confirming the validity of this approach.

We present a novel solution to the problem of scalable group key management. For the rekeying of the group key all members of the group should agree the rekeying of the key. However synchronization with all members for agreement of the rekeying is very costly. We have proposed an asynchronous rekeying framework based on the reliable and totally ordered multicast protocols (RTOMP). In our framework, a group consists of some domains which have some group members and one trusted key distribution server (KDS). The secure RTOMP channel is assumed for the communication channel among KDSs. When the membership of a domain is changed, the KDS of the domain creates the group key and distributes to the all other KDSs via the secure multicast channel on its own responsibility. Each member has an individual key between the KDS in the domain. The member can take a new key from the KDS in its own domain. Although this approach can disperse the cost of key creation and distribution to all KDSs, each KDS is responsible to key distribution to all members in the domain. Therefore, the scalability is limited by the performance of the KDS. We introduce the push-based key distribution and propose the combined approach of pull-based and push-based key distribution. From our simulation, the proposed approach can reduce the cost of the KDSs and improve the scalability.

In this paper, the theoretical studies and solutions regarding problems for realizing Soft TCCS (Time Critical Communication System), residing on top of Ethernet LAN are presented. We analyze Ethernet LAN including its precise back off sequence to obtain throughput and delay characteristics in a light loaded circumstances. We also analyze the same performance characteristics effected by multi-modal condition which is the mode of transmitting both short packets for control messages and long packets for bulk data concurrently in Factory Automation system. These theoretical studies are validated through computer simulations. We end by presenting the definition of the bandwidth in Ethernet LAN and also formulae regarding its estimation needed for admission control to maintain the total load within the predetermined value.

Noah (network oriented applications harmony) is a software platform to develop distributed systems that are flexible and reusable. Noah provides a flexible mechanism of communication by masking distribution aspects and synchronization between application modules. Noah has a "field" concept to manage not only communication but also cooperation among application modules. The paper describes Noah's basic concepts, architecture design, implementation, and application to office automation systems. The paper also presents Noah's integration mechanism for legacy applications.>

The authors describe the design of a testing environment, called FOREST (formal environment for systematic testing), for communication systems. The proposed method integrates three major strategies: (1) the use of formal description techniques (FDTs), for their formal semantics, which makes them unambiguous; (2) a stepwise approach to generate appropriate test cases for different test stages; and (3) the systematic support of the testing process, from test development to test execution. FOREST consists of four major subsystems. TENT (test sequence generation tool) is a subsystem that generates test cases from the formal specification of the behavior of a given protocol. APRICOT (ASN.1 pre-compiler, coder and test-debugger) is a subsystem that generates test data from the protocol data structures definition. TESPEC (test specification editor) is a tool producing some test specification based on a standardized formalism. The last subsystem, TEXEC (test executor), is a test execution system which consists of an upper tester and a lower tester, and a simulator of the communication medium.>

The testing environment FOREST (formal environment for systematic testing), which uses formal description techniques systematically, has been developed. FOREST generates the TTCN format test sequence from the specification described in SDL. Furthermore, FOREST generates the test data from the protocol data definition described in ASN.1. By using the FOREST system, the test specification and test execution and test report description are achieved systematically. Therefore, the cost of the testing phase of OSI upper layer software development is reduced. In particular, the facility for test sequence generation of FOREST realizes high coverage with a reasonable cost. The facility can efficiently check protocol specifications. The effectiveness of the approach is demonstrated by an application to the test design for the CCR protocol.>