Multi-person pose estimation is an important but challenging problem in computer vision. Although current approaches have achieved significant progress by fusing the multi-scale feature maps, they pay little attention to enhancing channel-wise and spatial information of maps. In this paper, we propose two novel modules perform enhancement for multi-person estimation. First, a Channel Shuffle Module (CSM) proposed adopt channel shuffle operation on maps with different levels, promoting...

Most existing facial landmark detection algorithms regard the manually annotated landmarks as precise hard labels, therefore, accurate are essential to training of these algorithms. However, in many cases, there exist deviations manual annotations, and marked for parts with occlusion large poses not always accurate, which means that “ground truth” usually precisely. In such case, it is more reasonable use soft labels rather than explicit labels. Therefore, this paper proposes associate a...

Natural disasters have caused substantial economic losses and numerous casualties. The demand analysis of relief supplies is the premise basis for efficient operations after disasters. With widespread use social media, it has become a vital channel people to report their provides way obtain information on disaster areas. Therefore, we present topic model-based framework establish dictionary gazetteer that aims identify spatial distribution by using media data. Taking 2013 Typhoon Haiyan...

Intelligent robotic systems have been extensively applied in factory automation, space exploration, intelligent buildings, surgery, military service, and also our daily life. Various remote control methods performed for systems, such as radio, microwave, computer networks, etc. Nowadays, the network services broadly used life, FTP, Telnet, World Wide Web, e-mail, Consequently, it is very convenient to use Internet robot, users will increase future. In past few years, many researchers using a...

Supervised learning is dominant in person search, but it requires elaborate labeling of bounding boxes and identities. Large-scale labeled training data often difficult to collect, especially for A natural question whether a good search model can be trained without the need identity supervision. In this paper, we present weakly supervised setting where only box annotations are available. Based on new setting, provide an effective baseline termed Region Siamese Networks (R-SiamNets). Towards...

To keep autonomous mobile robot in continuously working condition, power supply is an important issue. Typically, rechargeable batteries may provide few hours of peak usage. Recharging necessary before the has exhausted. In this paper, docking station with auto-recharging device and mechanism are implemented. We propose a control strategy automatic recharging for security recharging. also prediction algorithm to determine when needs navigate back An artificial landmark detected recognized by...

The security of home, laboratory, office and factory is essential to human daily life. A danger event often caused by the negligence humans. Potential hazards may injure our Therefore, it motivates us develop an intelligent multi-sensor based Security Robot system. It expected be widely employed in robot can detect dangerous situation provide timely alert us. structure contains eight parts. Including remote surveillance control system, image obstacle avoidance software auto-dialing, driver...

Dense Global Position System (GPS) arrays can be used to invert the terrestrial water-storage anomaly (TWSA) with higher accuracy. However, uneven distribution of GPS stations greatly limits application derive TWSA. Aiming solve this problem, we grid array using regression raise reliability TWSA inversion. First, study uses random forest (RF) model simulate crustal deformation in unobserved grids. Meanwhile, new Machine-Learning Loading-Inverted Method (MLLIM) is constructed based on...

Describes an algorithm and experimental results for detecting isolating sensory failures a fire detection system. In the system, we use two smoke sensors, flame sensors temperature to detect events. These can be classified into groups, each group has one sensor sensor. The failure isolation techniques described in paper are based on weight variation of adaptive fusion method. From simulation implementation results, it demonstrates that method exactly find out which is faulty isolate it. That...

Abstract Soil moisture (SM) is an essential Earth surface and climate system variable. Insights into its persistence corresponding time scales can improve numerical modeling prediction. In this study, SM from 17 observation stations within the Babao River Basin, Northwest China, between 1 June 31 August 2014 were used to investigate via adaptive fractal analysis. By conducting analysis of net radiation by estimating complementary cumulative distribution function precipitation intervals,...

In this paper, we propose a temporal feature augmented network for video instance segmentation. Video segmentation task can be split into two subtasks: and tracking. Similar to the previous work, track head is added an object instances across frames. Then performing detection, tracking tasks simultaneously. We choose Cascade-RCNN as basic network. Besides, in order make better use of rich information contained video, module introduced When on single frame, from other frames same will...

Multisensor fire detection algorithm (MSFDA) was one of the important elements in system for intelligent system. The paper proposed an adaptive fusion method detection, and uses smoke sensor, flame sensor temperature to detect a fire. In reality, phenomenon may have smoke, high temperature. But these signals can't happen at same time. First, we use computer simulation improve be adequate. Next, design modular by using ionization (TG-135), semiconductor (AD590) ultra-violet (R2868). We Taylor...

Based on 3D seismic data, the evolution mechanism and characteristics of faults were investigated to reveal structural origin its control differential hydrocarbon accumulation through comprehensive analyses, including structure style analysis, fault activity analogue modelling comparison among wells. The complex system with differently trending resulted from strike-slip rifting in Paleogene was partly activated, developed successively stretched obliquely by near-NS extensional stress field...

As the amount of social video content captured at physical-world events, and shared online, is rapidly increasing, there a growing need for robust methods organization presentation content. In this work, we significantly extend prior work that examined automatic detection videos from events were same time, i.e. "overlapping". We go beyond finding pairwise matches between clips describe construction scenes, or sets multiple overlapping videos, each scene presenting coherent moment in event....

The security of buildings, homes, laboratories, offices and factories is essential to human life. An unlucky event often caused by negligence. We have developed a multiagent multisensor-based system or intelligent building. can be widely employed in daily life detect dangerous situations using sensors. structure the divided into four parts, fire detection/diagnosis agent, intruder environment power agent. In this paper, we use an adaptive data fusion method agent rule-based statistical...

Multi-person pose estimation is an important but challenging problem in computer vision. Although current approaches have achieved significant progress by fusing the multi-scale feature maps, they pay little attention to enhancing channel-wise and spatial information of maps. In this paper, we propose two novel modules perform enhancement for multi-person estimation. First, a Channel Shuffle Module (CSM) proposed adopt channel shuffle operation on maps with different levels, promoting...

On-road risk detection is one of the most important applications Deep Neural Network (DNN). A good on-road system can accurately detect dangerous obstacles on road and provides drivers with early warning information to avoid danger. However, current high-precision systems usually use a cumbersome DNN, which require large number computing resources. Due limitations mobile devices such as limited power consumption problem, it difficult apply model for running real-time applications. Therefore...

The security system of home and building is an important issue to human being. We develop the low cost robot that widely employed in our daily life. can detect abnormal dangerous situation notify us thought Internet GSM module. structure contains seven parts. There are system, avoid obstacle driver software development auto recharging remote supervise multiple interface others. First, we a based using aluminum flame. Then present mobile (RMSS) for design interfaces Wu-Feng (WFSR-II)....