Objective Category-specific recognition and naming deficits have been observed in a variety of patient populations. However, the category-specific cortices for naming famous faces, animals and man-made objects remain ...Objective Category-specific recognition and naming deficits have been observed in a variety of patient populations. However, the category-specific cortices for naming famous faces, animals and man-made objects remain controversial. The present study aimed to study the specific areas involved in naming pictures of these 3 categories using functional magnetic resonance imaging. Methods Functional images were analyzed using statistical parametric mapping and the 3 different contrasts were evaluated using t statistics by comparing the naming tasks to their baselines.The contrast images were entered into a random-effects group level analysis.The results were reported in Montreal Neurological Institute co-ordinates,and anatomical regions were identified using an automated anatomical labeling method with XJview 8.Results Naming famous faces caused more activation in the bilateral head of the hippocampus and amygdala with significant left dominance. Bilateral activation of pars triangularis and pars opercularis in the naming of famous faces was also revealed. Naming animals evoked greater responses in the left supplementary motor area, while naming man-made objects evoked more in the left premotor area,left pars orbitalis and right supplementary motor area. The extent of bilateral fusiform gyri activation by naming man-made objects was much larger than that by naming of famous faces or animals.Even in the overlapping sites of activation,some differences among the categories were found for activation in the fusiform gyri.Conclusion The cortices involved in the naming process vary with the naming of famous faces,animals and man-made objects.This finding suggests that different categories of pictures should be used during intra-operative language mapping to generate a broader map of language function, in order to minimize the incidence of false-negative stimulation and permanent post-operative deficits.展开更多
Temporal action localization (TAL) is a task of detecting the start and end timestamps of action instances and classifying them in an untrimmed video. As the number of action categories per video increases, existing w...Temporal action localization (TAL) is a task of detecting the start and end timestamps of action instances and classifying them in an untrimmed video. As the number of action categories per video increases, existing weakly-supervised TAL (W-TAL) methods with only video-level labels cannot provide sufficient supervision. Single-frame supervision has attracted the interest of researchers. Existing paradigms model single-frame annotations from the perspective of video snippet sequences, neglect action discrimination of annotated frames, and do not pay sufficient attention to their correlations in the same category. Considering a category, the annotated frames exhibit distinctive appearance characteristics or clear action patterns.Thus, a novel method to enhance action discrimination via category-specific frame clustering for W-TAL is proposed. Specifically,the K-means clustering algorithm is employed to aggregate the annotated discriminative frames of the same category, which are regarded as exemplars to exhibit the characteristics of the action category. Then, the class activation scores are obtained by calculating the similarities between a frame and exemplars of various categories. Category-specific representation modeling can provide complimentary guidance to snippet sequence modeling in the mainline. As a result, a convex combination fusion mechanism is presented for annotated frames and snippet sequences to enhance the consistency properties of action discrimination,which can generate a robust class activation sequence for precise action classification and localization. Due to the supplementary guidance of action discriminative enhancement for video snippet sequences, our method outperforms existing single-frame annotation based methods. Experiments conducted on three datasets (THUMOS14, GTEA, and BEOID) show that our method achieves high localization performance compared with state-of-the-art methods.展开更多
Convolutional neural networks （CNNs） have been applied in state-of-the-art visual tracking tasks to represent the target. However, most existing algorithms treat visual tracking as an object-specific task. Therefore...Convolutional neural networks （CNNs） have been applied in state-of-the-art visual tracking tasks to represent the target. However, most existing algorithms treat visual tracking as an object-specific task. Therefore, the model needs to be retrained for different test video sequences. We propose a branch-activated multi-domain convolutional neural network （BAMDCNN）. In contrast to most existing trackers based on CNNs which require frequent online training, BAMDCNN only needs offine training and online fine-tuning. Specifically, BAMDCNN exploits category-specific features that are more robust against variations. To allow for learning category-specific information, we introduce a group algorithm and a branch activation method. Experimental results on challenging benchmark show that the proposed algorithm outperforms other state-of-the-art methods. What＇s more, compared with CNN based trackers, BAMDCNN increases tracking speed.展开更多
With the growing popularity of online social network, trust plays a more and more important role in connecting people to each other. We rely on our personal trust to accept recommendations, to make purchase decisions ...With the growing popularity of online social network, trust plays a more and more important role in connecting people to each other. We rely on our personal trust to accept recommendations, to make purchase decisions and to select transaction partners in the online community. Therefore, how to obtain trust relationships through mining online social networks becomes an important research topic. There are several shortcomings of existing trust mining methods. First, trust is category-dependent. However, most of the methods overlook the category attribute of trust relationships, which leads to low accuracy in trust calculation. Second, since the data in online social networks cannot be understood and processed by machines directly, traditional mining methods require much human effort and are not easily applied to other applications. To solve the above problems, we propose a semantic-based trust reasoning mechanism to mine trust relationships from online social networks automatically. We emphasize the category attribute of pairwise relationships and utilize Semantic Web technologies to build a domain ontology for data communication and knowledge sharing. We exploit role-based and behavior-based reasoning functions to infer implicit trust relationships and category-specific trust relationships. We make use of path expressions to extend reasoning rules so that the mining process can be done directly without much human effort. We perform experiments on real-life data extracted from Epinions. The experimental results verify the effectiveness and wide application use of our proposed method.展开更多
基金supported bythe Foundation of Science and Technology Program of Guangdong Province,China(No.2008A030201021)the Natural Science Foundation of Guangdong Province,China(No.10151001002000010)
文摘Objective Category-specific recognition and naming deficits have been observed in a variety of patient populations. However, the category-specific cortices for naming famous faces, animals and man-made objects remain controversial. The present study aimed to study the specific areas involved in naming pictures of these 3 categories using functional magnetic resonance imaging. Methods Functional images were analyzed using statistical parametric mapping and the 3 different contrasts were evaluated using t statistics by comparing the naming tasks to their baselines.The contrast images were entered into a random-effects group level analysis.The results were reported in Montreal Neurological Institute co-ordinates,and anatomical regions were identified using an automated anatomical labeling method with XJview 8.Results Naming famous faces caused more activation in the bilateral head of the hippocampus and amygdala with significant left dominance. Bilateral activation of pars triangularis and pars opercularis in the naming of famous faces was also revealed. Naming animals evoked greater responses in the left supplementary motor area, while naming man-made objects evoked more in the left premotor area,left pars orbitalis and right supplementary motor area. The extent of bilateral fusiform gyri activation by naming man-made objects was much larger than that by naming of famous faces or animals.Even in the overlapping sites of activation,some differences among the categories were found for activation in the fusiform gyri.Conclusion The cortices involved in the naming process vary with the naming of famous faces,animals and man-made objects.This finding suggests that different categories of pictures should be used during intra-operative language mapping to generate a broader map of language function, in order to minimize the incidence of false-negative stimulation and permanent post-operative deficits.
基金supported by the National Natural Science Foundation of China(No.61672268)。
文摘Temporal action localization (TAL) is a task of detecting the start and end timestamps of action instances and classifying them in an untrimmed video. As the number of action categories per video increases, existing weakly-supervised TAL (W-TAL) methods with only video-level labels cannot provide sufficient supervision. Single-frame supervision has attracted the interest of researchers. Existing paradigms model single-frame annotations from the perspective of video snippet sequences, neglect action discrimination of annotated frames, and do not pay sufficient attention to their correlations in the same category. Considering a category, the annotated frames exhibit distinctive appearance characteristics or clear action patterns.Thus, a novel method to enhance action discrimination via category-specific frame clustering for W-TAL is proposed. Specifically,the K-means clustering algorithm is employed to aggregate the annotated discriminative frames of the same category, which are regarded as exemplars to exhibit the characteristics of the action category. Then, the class activation scores are obtained by calculating the similarities between a frame and exemplars of various categories. Category-specific representation modeling can provide complimentary guidance to snippet sequence modeling in the mainline. As a result, a convex combination fusion mechanism is presented for annotated frames and snippet sequences to enhance the consistency properties of action discrimination,which can generate a robust class activation sequence for precise action classification and localization. Due to the supplementary guidance of action discriminative enhancement for video snippet sequences, our method outperforms existing single-frame annotation based methods. Experiments conducted on three datasets (THUMOS14, GTEA, and BEOID) show that our method achieves high localization performance compared with state-of-the-art methods.
基金the Innovation Action Plan Foundation of Shanghai(No.16511101200)
文摘Convolutional neural networks （CNNs） have been applied in state-of-the-art visual tracking tasks to represent the target. However, most existing algorithms treat visual tracking as an object-specific task. Therefore, the model needs to be retrained for different test video sequences. We propose a branch-activated multi-domain convolutional neural network （BAMDCNN）. In contrast to most existing trackers based on CNNs which require frequent online training, BAMDCNN only needs offine training and online fine-tuning. Specifically, BAMDCNN exploits category-specific features that are more robust against variations. To allow for learning category-specific information, we introduce a group algorithm and a branch activation method. Experimental results on challenging benchmark show that the proposed algorithm outperforms other state-of-the-art methods. What＇s more, compared with CNN based trackers, BAMDCNN increases tracking speed.
基金supported by the National Natural Science Foundation of China under Grant No. 61100183the Natural Science Foundation of Zhejiang Province of China under Grant No. Y1110477the Science Foundation of Zhejiang Sci-Tech University under Grant No. 0907838-Y
文摘With the growing popularity of online social network, trust plays a more and more important role in connecting people to each other. We rely on our personal trust to accept recommendations, to make purchase decisions and to select transaction partners in the online community. Therefore, how to obtain trust relationships through mining online social networks becomes an important research topic. There are several shortcomings of existing trust mining methods. First, trust is category-dependent. However, most of the methods overlook the category attribute of trust relationships, which leads to low accuracy in trust calculation. Second, since the data in online social networks cannot be understood and processed by machines directly, traditional mining methods require much human effort and are not easily applied to other applications. To solve the above problems, we propose a semantic-based trust reasoning mechanism to mine trust relationships from online social networks automatically. We emphasize the category attribute of pairwise relationships and utilize Semantic Web technologies to build a domain ontology for data communication and knowledge sharing. We exploit role-based and behavior-based reasoning functions to infer implicit trust relationships and category-specific trust relationships. We make use of path expressions to extend reasoning rules so that the mining process can be done directly without much human effort. We perform experiments on real-life data extracted from Epinions. The experimental results verify the effectiveness and wide application use of our proposed method.
