Person re-identification (reID) benefits greatly from deep convolutional neural networks (CNNs) which learn robust feature embeddings. However, CNNs are inherently limited in modeling the large variations in person po...
详细信息
ISBN:
(纸本)9781728132945
Person re-identification (reID) benefits greatly from deep convolutional neural networks (CNNs) which learn robust feature embeddings. However, CNNs are inherently limited in modeling the large variations in person pose and scale due to their fixed geometric structures. In this paper, we propose a novel network structure, Interaction-and-Aggregation (IA), to enhance the feature represen tation capability of CNNs. Firstly, Spatial IA (SIA) module is introduced. It models the interdependencies between spatial features and then aggregates the correlated features corresponding to the same body parts. Unlike CNNs which extract features from fixed rectangle regions, SIA can adaptively determine the receptive fields according to the input person pose and scale. Secondly, we introduce Channel IA (CIA) module which selectively aggregates channel features to enhance the feature represen tation, especially for small-scale visual cues. Further, IA network can be constructed by inserting IA blocks into CNNs at any depth. We validate the effectiveness of our model for person reID by demonstrating its superiority over state-of-the-art methods on three benchmark datasets.
Semantic Web (SW) has attracted the increasing attention of researchers, which facilitates people to link and handle various data. Ontology is the kernel technique of SW, and biomedical ontology is a state-of-art biom...
详细信息
ISBN:
(数字)9781728162157
ISBN:
(纸本)9781728162164
Semantic Web (SW) has attracted the increasing attention of researchers, which facilitates people to link and handle various data. Ontology is the kernel technique of SW, and biomedical ontology is a state-of-art biomedical knowledge modeling technique, which formally defines the biomedical concepts and their relationships. However, the same biomedical concepts in different biomedical ontologies could be defined in various contexts or with different terms, which yields the biomedical ontology heterogeneity problem. It is crucial to find mapping among heterogeneity concepts of different biomedical ontologies for bridging the semantic gaps, which is the so-called biomedical ontology matching. Biomedical ontology matching problem is an open challenge due to the rich semantic meaning and the flexible representation on a biomedical concept. To address this challenging problem, in this work, it is regarded as a binary classification problem, and a Long Short-Term Memory Networks (LSTM)-based ontology matching technique is proposed to solve it. Our proposal improves the quality of the alignment by introducing the char-embedding technique, which takes into account the semantic and context information of concepts. The comparing results with OAEI's participants show the effectiveness of our proposal.
The paper discusses fuzzy comparison matrices, consistency check, weight prioritization methods and weight evaluation methods in fuzzy group analytic hierarchy process. There are various methods of weight prioritizati...
详细信息
The paper discusses fuzzy comparison matrices, consistency check, weight prioritization methods and weight evaluation methods in fuzzy group analytic hierarchy process. There are various methods of weight prioritization, however, they are not critically evaluated. In the paper, two measures are introduced for the evaluation of the group weights. Then, a new method is proposed to improve the process of deriving weights and use it in an application compared to another common three methods. Our results show that the new method is a good method for deriving weights of indexes.
In practice, there often exist some occasions where video surveillance can only be realized by using rechargeable batteries due to the high cost of power supply. In order to extend the battery life, images can only be...
详细信息
ISBN:
(纸本)9781728105512
In practice, there often exist some occasions where video surveillance can only be realized by using rechargeable batteries due to the high cost of power supply. In order to extend the battery life, images can only be taken at a certain interval. Obviously, image difference is an important method to obtain various changes in this case. Among these changes, foreign invasion is one of the focuses. However, due to bolt shedding, abnormal weather and other factors, the camera may have abnormal movement which may seriously affect the detection effect of image difference method. Therefore, it is an urgent issue to find an effective way to detect the abnormal movement of camera and improve the detection accuracy of foreign invasion. Considering the characteristics of this kind of image sequences, we propose an effective algorithm to detect the camera abnormal movement and foreign object invasion based on a cumulative edge distribution probability model. Since sky region is relatively simple, we only discuss changes in sky region to detect camera abnormal movement. Our algorithm has 6 basic steps: firstly, segment the sky region;secondly, extract the edge information of the current image and the preceding adjacent image in the image sequence;thirdly, determine if the edge information of two adjacent images coincide. If consistent, then go to the next step, otherwise, it indicates that the camera has abnormal movement, then alarm;fourthly, calculate the cumulative edge probability distribution model in the sky region by using the historical image sequence;fifthly, by using adaptive Parzen window, determine if foreign object invasion exists by comparing the probability model of cumulative edge distribution with edge distribution of the current image;sixthly, update the cumulative edge distribution probability model in sky region. The algorithm achieves good results in practical applications. Through the test of thousands of images taken in the wild, the detection accuracy of
Video person re-identification (re-ID) plays an important role in surveillance video analysis. However, the performance of video re-ID degenerates severely under partial occlusion. In this paper, we propose a novel ne...
详细信息
ISBN:
(纸本)9781728132945
Video person re-identification (re-ID) plays an important role in surveillance video analysis. However, the performance of video re-ID degenerates severely under partial occlusion. In this paper, we propose a novel network, called Spatio-Temporal Completion network (STCnet), to explicitly handle partial occlusion problem. Different from most previous works that discard the occluded frames, STCnet can recover the appearance of the occluded parts. For one thing, the spatial structure of a pedestrian frame can be used to predict the occluded body parts from the unoccluded body parts of this frame. For another, the temporal patterns of pedestrian sequence provide important clues to generate the contents of occluded parts. With the spatio-temporal information, STCnet can recover the appearance for the occluded parts, which could be leveraged with those unoccluded parts for more accurate video re-ID. By combining a re-ID network with STCnet, a video re-ID framework robust to partial occlusion (VRSTC) is proposed. Experiments on three challenging video re-ID databases demonstrate that the proposed approach outperforms the state-of-the-arts.
Chinese is a logographic writing system, and the shape of Chinese characters contain rich syntactic and semantic information. In this paper, we propose a model to learn Chinese word embeddings via three-level composit...
详细信息
作者：
Ende WangZhiyuan LiuBing WangZhiyu CaoShiwei Zhanga Key Laboratory of Opto-Electronic Information Processing
Chinese Academy of Sciences Shenyang People’s Republic of Chinab Shenyang Institute of Automation Chinese Academy of Sciences Shenyang People’s Republic of Chinac Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences Shenyang People’s Republic of Chinad University of Chinese Academy of Sciences Beijing People's Republic of China a Key Laboratory of Opto-Electronic Information Processing
Chinese Academy of Sciences Shenyang People’s Republic of Chinab Shenyang Institute of Automation Chinese Academy of Sciences Shenyang People’s Republic of Chinac Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences Shenyang People’s Republic of Chinae School of Automation and Electrical Engineering Shenyang Ligong University Shenyang People's Republic of China f AVIC Hongdu Aviation Industry Group
Nanchang People’s Republic of China a Key Laboratory of Opto-Electronic Information Processing
Chinese Academy of Sciences Shenyang People’s Republic of Chinab Shenyang Institute of Automation Chinese Academy of Sciences Shenyang People’s Republic of Chinac Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences Shenyang People’s Republic of Chinag College of Information Liaoning University Shenyang People’s Republic of China
Infrared image systems often generate stripe noise because of the non-uniformity of the focal plane array, which significantly reduces the visual quality of the image. To solve this problem, we proposed an effective s...
详细信息
Infrared image systems often generate stripe noise because of the non-uniformity of the focal plane array, which significantly reduces the visual quality of the image. To solve this problem, we proposed an effective single-frame denoising method in this paper. First, the wavelet function extracts the approximate and vertical components of the original image containing stripe noise. Then, the approximation component is denoised by parameter estimation, and the vertical component is denoised by guided filtering. Finally, wavelet reconstruction is performed to realize the denoising process of the original image. This method avoids the loss of details of other components of the image and achieves an excellent denoising effect. The experimental results on the public datasets indicate that our proposed method can effectively eliminate the stripe noise of infrared images compared with some advanced methods.
Aspect-based sentiment analysis (ABSA), which aims to identify fine-grained opinion polarity towards a specific aspect, is a challenging subtask of sentiment analysis (SA). In this paper, we construct an auxiliary sen...
详细信息
Recommender systems are one of the most important technologies in the electronic commerce system. In a collaborative filtering recommendation algorithm, similarity calculation is the key to determining the efficiency ...
Recommender systems are one of the most important technologies in the electronic commerce system. In a collaborative filtering recommendation algorithm, similarity calculation is the key to determining the efficiency of the recommendation algorithm. This paper analyzes the shortcomings of traditional similarity measurement methods in recommender systems and proposes a scoring-matrix-filling algorithm. Based on information categories and user interest similarity, the algorithm can reduce the negative influence of data sparsity on the recommendation result to some extent. The research results have certain practical and guiding significance.
Different linguistic perspectives cause many diverse segmentation criteria for Chinese word segmentation (CWS). Most existing methods focus on improving the performance of single-criterion CWS. However, it is interest...
详细信息
暂无评论