Today massive collections of data can be obtained across different sources (or domains), e.g., the depth data from Kinect, the geometrical data from scanning devices, the imagery/video data from cameras, and the motio...Today massive collections of data can be obtained across different sources (or domains), e.g., the depth data from Kinect, the geometrical data from scanning devices, the imagery/video data from cameras, and the motion data from mocap devices. Since heterogeneous data may have different discriminative powers and are intrinsically complementary for certain tasks, it is desirable to leverage all展开更多
Question answering is an important problem that aims to deliver specific answers to questions posed by humans in natural language.How to efficiently identify the exact answer with respect to a given question has becom...Question answering is an important problem that aims to deliver specific answers to questions posed by humans in natural language.How to efficiently identify the exact answer with respect to a given question has become an active line of research.Previous approaches in factoid question answering tasks typically focus on modeling the semantic relevance or syntactic relationship between a given question and its corresponding answer.Most of these models suffer when a question contains very little content that is indicative of the answer.In this paper,we devise an architecture named the temporality-enhanced knowledge memory network(TE-KMN) and apply the model to a factoid question answering dataset from a trivia competition called quiz bowl.Unlike most of the existing approaches,our model encodes not only the content of questions and answers,but also the temporal cues in a sequence of ordered sentences which gradually remark the answer.Moreover,our model collaboratively uses external knowledge for a better understanding of a given question.The experimental results demonstrate that our method achieves better performance than several state-of-the-art methods.展开更多
Images with human faces comprise an essential part in the imaging realm. Occlusion or damage in facial portions will bring a remarkable discomfort and information loss. We propose an algorithm that can repair occluded...Images with human faces comprise an essential part in the imaging realm. Occlusion or damage in facial portions will bring a remarkable discomfort and information loss. We propose an algorithm that can repair occluded or damaged facial images automatically,named 'facial image inpainting'. Inpainting is a set of image processing methods to recover missing image portions. We extend the image inpainting methods by introducing facial domain knowledge. With the support of a face database,our ap-proach propagates structural information,i.e.,feature points and edge maps,from similar faces to the missing facial regions. Using the inferred structural information as guidance,an exemplar-based image inpainting algorithm is employed to copy patches in the same face from the source portion to the missing portion. This newly proposed concept of facial image inpainting outperforms the traditional inpainting methods by propagating the facial shapes from a face database,and avoids the problem of variations in imaging conditions from different images by inferring colors and textures from the same face image. Our system produces seamless faces that are hardly seen drawbacks.展开更多
We present a complete framework for synthesizing style-preserving 2D cartoons by learning from traditional Chinese cartoons. In contrast to reusing-based approaches which rely on rearranging or retrieving existing car...We present a complete framework for synthesizing style-preserving 2D cartoons by learning from traditional Chinese cartoons. In contrast to reusing-based approaches which rely on rearranging or retrieving existing cartoon sequences, we aim to generate stylized cartoons with the idea of style factorization. Specifically, starting with 2D skeleton features of cartoon characters extracted by an improved rotoscoping system, we present a non-negative style factorization (NNSF) algorithm to obtain style basis and weights and simultaneously preserve class separability. Thus, factorized style basis can be combined with heterogeneous weights to re-synthesize style-preserving features, and then these features are used as the driving source in the character reshaping process via our proposed subkey-driving strategy. Extensive experiments and examples demonstrate the effectiveness of the proposed framework.展开更多
文摘Today massive collections of data can be obtained across different sources (or domains), e.g., the depth data from Kinect, the geometrical data from scanning devices, the imagery/video data from cameras, and the motion data from mocap devices. Since heterogeneous data may have different discriminative powers and are intrinsically complementary for certain tasks, it is desirable to leverage all
基金supported by the National Basic Research Program(973)of China(No.2015CB352302)the National Natural Science Foundation of China(Nos.61625107,U1611461,U1509206,and 61402403)+2 种基金the Key Program of Zhejiang Province,China(No.2015C01027)the Chinese Knowledge Center for Engineering Sciences and Technologythe Fundamental Research Funds for the Central Universities,China
文摘Question answering is an important problem that aims to deliver specific answers to questions posed by humans in natural language.How to efficiently identify the exact answer with respect to a given question has become an active line of research.Previous approaches in factoid question answering tasks typically focus on modeling the semantic relevance or syntactic relationship between a given question and its corresponding answer.Most of these models suffer when a question contains very little content that is indicative of the answer.In this paper,we devise an architecture named the temporality-enhanced knowledge memory network(TE-KMN) and apply the model to a factoid question answering dataset from a trivia competition called quiz bowl.Unlike most of the existing approaches,our model encodes not only the content of questions and answers,but also the temporal cues in a sequence of ordered sentences which gradually remark the answer.Moreover,our model collaboratively uses external knowledge for a better understanding of a given question.The experimental results demonstrate that our method achieves better performance than several state-of-the-art methods.
基金supported by the National Natural Science Foundation of China (No. 60525108)the National Key Technology R & D Program of China (No. 2006BAH11B03-4)
文摘Images with human faces comprise an essential part in the imaging realm. Occlusion or damage in facial portions will bring a remarkable discomfort and information loss. We propose an algorithm that can repair occluded or damaged facial images automatically,named 'facial image inpainting'. Inpainting is a set of image processing methods to recover missing image portions. We extend the image inpainting methods by introducing facial domain knowledge. With the support of a face database,our ap-proach propagates structural information,i.e.,feature points and edge maps,from similar faces to the missing facial regions. Using the inferred structural information as guidance,an exemplar-based image inpainting algorithm is employed to copy patches in the same face from the source portion to the missing portion. This newly proposed concept of facial image inpainting outperforms the traditional inpainting methods by propagating the facial shapes from a face database,and avoids the problem of variations in imaging conditions from different images by inferring colors and textures from the same face image. Our system produces seamless faces that are hardly seen drawbacks.
基金supported by the National Basic Research Program of China(No.2015CB352300)the National Natural Science Foundation of China(Nos.61402401 and U1509206)+3 种基金the Zhejiang Provincial Natural Science Foundation of China(No.LQ14F010004)the China Knowledge Centre for Engineering Sciences and Technologythe Fundamental Research Funds for the Central Universitiesthe Qianjiang Talents Program of Zhejiang Province,China
基金supported by the National Basic Research Program (973) of China (No. 2012CB316400)the National Natural Science Foundation of China (No. 60903134)the Natural Science Foundation of Zhejiang Province, China (No. Y1101129)
文摘We present a complete framework for synthesizing style-preserving 2D cartoons by learning from traditional Chinese cartoons. In contrast to reusing-based approaches which rely on rearranging or retrieving existing cartoon sequences, we aim to generate stylized cartoons with the idea of style factorization. Specifically, starting with 2D skeleton features of cartoon characters extracted by an improved rotoscoping system, we present a non-negative style factorization (NNSF) algorithm to obtain style basis and weights and simultaneously preserve class separability. Thus, factorized style basis can be combined with heterogeneous weights to re-synthesize style-preserving features, and then these features are used as the driving source in the character reshaping process via our proposed subkey-driving strategy. Extensive experiments and examples demonstrate the effectiveness of the proposed framework.