Intelligent Media for Realizing Question-Answering with Appropriate Images and Videos

智能媒体，通过合适的图片和视频实现问答

基本信息

批准号：
13480087
负责人：
NAKAMURA Yuichi
金额：
$ 2.56万
依托单位：
University of Tsukuba
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B)
财政年份：
2001
资助国家：
日本
起止时间：
2001 至 2003
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-13480087/
关键词：
multimedia video-based media interactive media video processing natural language processing communication media intelligent video capturing automated editing

项目摘要

The purpose of this project is to realize an intelligent interactive media that answers questions based on video-based media.. Through three year's investigation, we obtained the following results・We showed automated video capturing and indexing are efficiently integrated in our system. The objects handled in a. presentation and the manipulation done for the objects are effectively detected lay our multi-sensor system・We proposed a method for automatic video indexcing for broadcasted contents such as cooking shows・We proposed a framework QUEVICO, which is a question-based video composition scheme in which video indexing and editing is designed from the viewpoint of "question find answer", and in which multi-view videos can be effectively used. Based on this framework, the system retrieves a relevant video portion, even if a complete set of indices ire not given・We proposed a novel model for video editing and summarizing based on constraint-satisfaction. In this model, a video is considered as a sequence of boxes i.e., video segments, each of which has a short length, e.g., 1 second, and the editing is defined as the problem of assigning an appropriate shot to each box. A variety of good editing patterns can be generated, and that we can control the characteristics of obtained videos by changing the rules and parameters

该项目的目的是实现一个智能的互动媒体，该媒体根据基于视频的媒体回答问题。通过三年的调查，我们获得了以下结果。我们显示自动化视频捕获和索引有效地集成在我们的系统中。对象在A中处理。有效地检测到了对象的演示和对物体进行的操作，我们提出了一种用于自动视频索引的方法，用于播放的内容，例如烹饪，例如烹饪显示・我们提出了一个基于问题的视频组成方案，这是一种基于问题的视频组成方案，从哪个问题上设计了“构图和编辑”的“问题”的“问题”的“求解”。基于此框架，系统会检索相关的视频部分，即使未给出的一组完整的索引也没有给出・我们提出了一种基于约束满意度的视频编辑和汇总的新颖模型。在此模型中，视频被认为是一系列框，即视频片段，每个片段的长度很短，例如1秒，并且编辑被定义为向每个盒子分配适当的镜头的问题。可以生成各种良好的编辑模式，我们可以通过更改规则和参数来控制获得视频的特征