| Summary: | 碩士 === 國立政治大學 === 資訊科學學系 === 101 === With the advent of the digital era, there is an exponential growth of digital photos and images. By combining photo memories and interactive storytelling, we can not only reduce the amount of time needed for exhaustively browsing large photo collections, but also help a user narrate a story and value up the existing photos. We observe that most of previous photo-based researches focus on the topics of automatically generating photo tags and doing image analysis, but rarely dedicate their efforts to photo-based interactive storytelling. In this work, a creation platform for narrating stories based on memories and photos is proposed. The system allows a user to reminisce the past and stimulates creative thinking during the process of story creation, which turns traditional photo browsing into a totally different and pleasant experience. In the proposed platform, keywords annotated in photos are used for similarity comparison, and similar photos are then extracted for serving the templates of narrative memory. A new story- or event-based photo album is eventually generated based on the interaction between the system and the user. Therefore, every use of the system is a whole new experience for a user since the system always dynamically delivers related content to users according to their responses and feedbacks.
We invited users to participate in the platform experiment with different content delivery methods, including 1) random, 2) static, and 3) dynamic, to verify the feasibility of the proposed algorithm. Based on the interviews and questionnaires collected from the users, a quantitative analysis is conducted to verify the effectiveness and usability of the proposed platform. Through the experiment, we observe that the proposed system can evidently help user in story creation under the setting of the keyword-based story layout and the method of dynamically adjustive selection. The reason could be that an explicit story layout can be relatively easier to acquire user's recognition, and the photos suggested by the proposed platform can better match user's expectation. In contrast, time and location information in a photo can only be used for supplementary retrieval filter only.
|
|---|
