A VIDEO-SEQUENCE AUTOMATICAL CONTINUATION METHOD BASED ON THE VIDEO-SCENES VISUAL AND SEMANTIC PROPERTIES

Author(s): Vishnevskaya T.I., Utochkina N.V.

Rubric: Information technology

DOI: 10.21777/2500-2112-2021-3-53-59

Release: 2021-3 (36)

Pages: 53-59

Keywords: recommendations, video, video scene properties, hierarchical RNN, video-sequence continuation

Annotation: This article suggests a continuation method for a video-sequence by means of hierarchical recurrent neural networks. The method is based on the information about the visual and semantic properties of video-scenes obtained fr om the video and its previously known data. The article presents algorithms for identifying the visual and semantic properties of video- scenes and an algorithm for extending the video sequence as well. It describes the result of the method and demonstrates its worth. It also describes possible scenarios its usage. The method can be used for recommending short video clips and films. As a result a set of videos is obtained wh ere each element is close to the initial selection in terms of the selected properties.

Bibliography: Vishnevskaya T.I., Utochkina N.V. A VIDEO-SEQUENCE AUTOMATICAL CONTINUATION METHOD BASED ON THE VIDEO-SCENES VISUAL AND SEMANTIC PROPERTIES // Education Resources and Technologies. – 2021. – № 3 (36). – С. 53-59. doi: 10.21777/2500-2112-2021-3-53-59