Đang chuẩn bị liên kết để tải về tài liệu:
The Inverse Hollywood Problem
 From video to scripts and storyboards via causal analysis

The widespread appeal of Hollywood movies is due not only to the studios’ economic power and marketing prowess but also to the formal-aesthetic qualities of the films themselves. This third aspect of the term Hollywood has changed somewhat less than the industrial and institutional aspects, in that the cinematic style and narrative struc- ture of Hollywood movies have persisted over the decades, despite the obvious need for novelty and innovation. In other words, what we call a “Hollywood movie” is much the same artifact today as it was in the late teens and early 1920s. Recent changes in Hollywood’s industrial and institutional operations threaten this formal-aesthetic stability, however, due to. | MIT Media Lab Perceptual Computing Learning and Common Sense Technical Report 410 28dec96 The Inverse Hollywood Problem From video to scripts and storyboards via causal analysis Matthew Brand The Media Lab MIT 20 Ames Street Cambridge MA 02139 USA brand media.mit.edu WWW.media.mit.edu brand Abstract We address the problem of visually detecting causal events and fitting them together into a coherent story of the action witnessed by the camera. We show that this can be done by reasoning about the motions and collisions of surfaces using high-level causal constraints derived from psychological studies of infant visual behavior. These constraints are naive forms of basic physical laws governing substantiality contiguity momentum and acceleration. We describe two implementations. One system parses instructional videos extracting plans of action and key frames suitable for storyboarding. Since learning will play a role in making such systems robust we introduce a new framework for coupling hidden Markov models and demonstrate its use in a second system that segments stereo video into actions in near real-time. Rather than attempt accurate low-level vision both systems use high-level causal analysis to integrate fast but sloppy pixel-based representations over time. The output is suitable for summary indexing and automated editing. @1997 AAAI. All rights reserved. Introduction A useful result from a vision system would be an answer to the question What is happening This is a question about causality What are the events and how do earlier ones cause or enable later ones We are exploring the hypothesis that causal perception rests on inference about the motions and collisions of surfaces and proceeds independently of processes such as recognition reconstruction and static segmentation . In this paper we present two computational models of this process one heuristic one probabilistic and trainable that incorporate psychological models of causal event perception in infants.