Academia.eduAcademia.edu

Figure 40 – uploaded by Rodrigo Laiola Guimaraes

See full PDFdownloadDownload figure
Figure 4.12. Results of the questionnaires in Woodbridge (UK).

Figure 4 12. Results of the questionnaires in Woodbridge (UK).

Related Figures (52)

Figure 1.1. From the authoring perspective, the main challenge has been to make content creation a manageable process.
Figure 1.1. From the authoring perspective, the main challenge has been to make content creation a manageable process.
Figure 1.3. High-level architecture of socially-aware multimedia authoring systems.
Figure 1.3. High-level architecture of socially-aware multimedia authoring systems.
The research question introduced above puts the accent on the social aspects around personal media. While knowledge from online social networks could be mined to determine the strength of ties among people [24], user interest and sentiment analysis also could be used to facilitate media annotation and content
The research question introduced above puts the accent on the social aspects around personal media. While knowledge from online social networks could be mined to determine the strength of ties among people [24], user interest and sentiment analysis also could be used to facilitate media annotation and content
Figure 2.1. Overview of the requirements and validation parameters for socially-aware multimedia authoring systems.
Figure 2.1. Overview of the requirements and validation parameters for socially-aware multimedia authoring systems.
Table 2.2. Data gathering events. emporal a emporal a perceptual he tempora be provided state-of-the common time base analysis, we will no ignmen ignment algorithm and a Semantic Vid tool is used to align all of the ime-frequency analysis with a precision o 1 alignment of a recorded dataset (more informa’ below). The level of accuracy of our tool i art solutions [44][59]. Since the focus of this further detail this part of the system. T find the a reported in boundaries, gorithm and its evaluation elsewhere [20]. [59]. The tool is capable of automatically d of fragmenting the clips into coherent units, resulting video sub-clips. In the next sections. we discuss the media gat s of around 99% ion on the d hesis is not he interested eo Annotation Suite. The individual video clips to a . The core of the temporal alignment algorithm is based on f 10ms. Figure 2.2 sketches atasets will , improving on content reader can The Semantic Video Annotation Suite [64] provides basic analysis functions, similar to the ones etecting po hering and ential shot and of annotating the annotation
Table 2.2. Data gathering events. emporal a emporal a perceptual he tempora be provided state-of-the common time base analysis, we will no ignmen ignment algorithm and a Semantic Vid tool is used to align all of the ime-frequency analysis with a precision o 1 alignment of a recorded dataset (more informa’ below). The level of accuracy of our tool i art solutions [44][59]. Since the focus of this further detail this part of the system. T find the a reported in boundaries, gorithm and its evaluation elsewhere [20]. [59]. The tool is capable of automatically d of fragmenting the clips into coherent units, resulting video sub-clips. In the next sections. we discuss the media gat s of around 99% ion on the d hesis is not he interested eo Annotation Suite. The individual video clips to a . The core of the temporal alignment algorithm is based on f 10ms. Figure 2.2 sketches atasets will , improving on content reader can The Semantic Video Annotation Suite [64] provides basic analysis functions, similar to the ones etecting po hering and ential shot and of annotating the annotation
Chapter 2. Personalized Memories of Social Events: Studying Asynchronous Togetherness
Chapter 2. Personalized Memories of Social Events: Studying Asynchronous Togetherness
Figure 2.3. The makeup, age and gender of participants in phase | evaluation.
Figure 2.3. The makeup, age and gender of participants in phase | evaluation.
Figure 2.4. The makeup, age and gender of participants in phase 2 evaluation. Chapter 2. Personalized Memories of Social Events: Studying Asynchronous Togetherness used this session as an opportunity to understand how participants conceptualized the concert. The second (in-depth) session was dedicated to capture video editing practices and media sharing routines of the participants, based on their interactions with the system. We used the footage they had recorded during the high school concert in the spring of 2010 to evaluate our initial prototype system. Both sessions were started with an ice-breaking activity on the whiteboard, followed by discussions around the research questions.
Figure 2.4. The makeup, age and gender of participants in phase 2 evaluation. Chapter 2. Personalized Memories of Social Events: Studying Asynchronous Togetherness used this session as an opportunity to understand how participants conceptualized the concert. The second (in-depth) session was dedicated to capture video editing practices and media sharing routines of the participants, based on their interactions with the system. We used the footage they had recorded during the high school concert in the spring of 2010 to evaluate our initial prototype system. Both sessions were started with an ice-breaking activity on the whiteboard, followed by discussions around the research questions.
Figure 2.5. A schematic view of the perceived strength of social bond over time in relation to our scenario.
Figure 2.5. A schematic view of the perceived strength of social bond over time in relation to our scenario.
b) Media sharing habits and social relations. Generic Architecture for Socially-Aware Authoring Systems Figure 2.6. Results of the questionnaires about social practice: around personal videos (phase 1 evaluation).
b) Media sharing habits and social relations. Generic Architecture for Socially-Aware Authoring Systems Figure 2.6. Results of the questionnaires about social practice: around personal videos (phase 1 evaluation).
1) Video recording and editing practices.
1) Video recording and editing practices.
Figure 2.7. Social media habits (phase 2 evaluation). Chapter 2. Personalized Memories of Social Events: Studying Asynchronous Togetherness
Figure 2.7. Social media habits (phase 2 evaluation). Chapter 2. Personalized Memories of Social Events: Studying Asynchronous Togetherness
Figure 2.8. Comparison between automatic and manual generation of video compilations. Automatic methods are not sufficient for creating personal and intimate memories.
Figure 2.8. Comparison between automatic and manual generation of video compilations. Automatic methods are not sufficient for creating personal and intimate memories.
Figure 2.9. Utility and usefulness of My Videos. Results of the questionnaire from phase 1 evaluation (Amsterdam/NL).
Figure 2.9. Utility and usefulness of My Videos. Results of the questionnaire from phase 1 evaluation (Amsterdam/NL).
Figure 2.10. Utility and usefulness of MyVideos. Results of the questionnaire from phase 2 evaluation (Woodbridge/UK).
Figure 2.10. Utility and usefulness of MyVideos. Results of the questionnaire from phase 2 evaluation (Woodbridge/UK).
Figure 3.1. Typical interface for watching videos on the Web. It does not take into account the social affinity between viewers and subjects featured in the video.
Figure 3.1. Typical interface for watching videos on the Web. It does not take into account the social affinity between viewers and subjects featured in the video.
Figure 3.2. Initial prototype implementation for browsing videos (thumbnail view).
Figure 3.2. Initial prototype implementation for browsing videos (thumbnail view).
Figure 3.3. Initial prototype implementation for browsing videos (timeline view). a i i I I To address these limitations of current video sharing services, our initial video browsing interface offered two views for exploring community contributed video clips. The thumbnail view (Figure 3.2) displayed media assets in a paginated grid, while the timeline view (Figure 3.3) showed how recorded videos temporally fitted the event timeline. In both views a user could apply six different filters to refine a query. These filters were: all media, my media, cameras, people, instruments and events. ‘All media’ referred to all videos uploaded to the system. ‘My media’ restricted navigation to only the videos uploaded by the current user. ‘Cameras’, ‘people’, ‘instruments’, and ‘events’ filters would display the respective annotated video clips based on the filter selection (e.g., ‘Julia’ or ‘Drums’).
Figure 3.3. Initial prototype implementation for browsing videos (timeline view). a i i I I To address these limitations of current video sharing services, our initial video browsing interface offered two views for exploring community contributed video clips. The thumbnail view (Figure 3.2) displayed media assets in a paginated grid, while the timeline view (Figure 3.3) showed how recorded videos temporally fitted the event timeline. In both views a user could apply six different filters to refine a query. These filters were: all media, my media, cameras, people, instruments and events. ‘All media’ referred to all videos uploaded to the system. ‘My media’ restricted navigation to only the videos uploaded by the current user. ‘Cameras’, ‘people’, ‘instruments’, and ‘events’ filters would display the respective annotated video clips based on the filter selection (e.g., ‘Julia’ or ‘Drums’).
Figure 3.4. Initial interface for adding and correcting annotations. Chapter 3. Designing Socially-Aware Video Exploration from Community Assets
Figure 3.4. Initial interface for adding and correcting annotations. Chapter 3. Designing Socially-Aware Video Exploration from Community Assets
chi im During the evaluation session, participants were actively looking mediately share video c ips with members of their close circle. “Can now?” was a common reaction after seeing a video clip they especia WwW rat her download the video video by email or share on indicated that a ‘Burn to D convenient given that grand files to their local computers, send a link of a YouTube and/or Facebook. Parents, on the ot for video ps of their close friends and relatives. In particular, some participants wanted to I send it ly liked. hen asked how they would share the videos, teenagers expressed they would particular her hand, VD’ functionality of the selected videos also would be parents usually do not have Internet access at home. When nromnted ahont what they remembhered of the concert moct
chi im During the evaluation session, participants were actively looking mediately share video c ips with members of their close circle. “Can now?” was a common reaction after seeing a video clip they especia WwW rat her download the video video by email or share on indicated that a ‘Burn to D convenient given that grand files to their local computers, send a link of a YouTube and/or Facebook. Parents, on the ot for video ps of their close friends and relatives. In particular, some participants wanted to I send it ly liked. hen asked how they would share the videos, teenagers expressed they would particular her hand, VD’ functionality of the selected videos also would be parents usually do not have Internet access at home. When nromnted ahont what they remembhered of the concert moct
mentioned that it would be interesting to have a spatial representation of the concert venue to help browsing the event footage. When inquired about particular events they remembered, participants reported on solos performed by different musicians. Among the youngest participants, an event in particular was pointed out as the most memorable of the concert.
mentioned that it would be interesting to have a spatial representation of the concert venue to help browsing the event footage. When inquired about particular events they remembered, participants reported on solos performed by different musicians. Among the youngest participants, an event in particular was pointed out as the most memorable of the concert.
Figure 3.8. Browsing interface based on the concert program. Chapter 3. Designing Socially-Aware Video Exploration from Community Assets
Figure 3.8. Browsing interface based on the concert program. Chapter 3. Designing Socially-Aware Video Exploration from Community Assets
Figure 3.9. Paper-based concert program handed out at the day of the event. second requirement. Hovering the mouse over interface elements (i.e., performers, songs and clips) also provides efficient and informative feedback. For instance, when a user hovers the mouse cursor over a performer thumbnail, the associated songs and media clips containing that person are highlighted in the user interface. This functionality, which has been designed to react in rapid response time, reduces the short-term memory load [47] and makes clear the relationship between performers, songs and clips.
Figure 3.9. Paper-based concert program handed out at the day of the event. second requirement. Hovering the mouse over interface elements (i.e., performers, songs and clips) also provides efficient and informative feedback. For instance, when a user hovers the mouse cursor over a performer thumbnail, the associated songs and media clips containing that person are highlighted in the user interface. This functionality, which has been designed to react in rapid response time, reduces the short-term memory load [47] and makes clear the relationship between performers, songs and clips.
Chapter 3. Designing Socially-Aware Video Exploration from Community Assets
Chapter 3. Designing Socially-Aware Video Exploration from Community Assets
While exploring videos displayed in the concert program, users expressed that they were having an engaging experience, but they did not have an option to play a song from begin to end. This feedback suggests the need for supporting more complex narrative alternatives that not only take into account the temporal alignment between videos, but also the preferences and social relationships of each individual user. This subject is the focus of the next chapter, which discusses the
While exploring videos displayed in the concert program, users expressed that they were having an engaging experience, but they did not have an option to play a song from begin to end. This feedback suggests the need for supporting more complex narrative alternatives that not only take into account the temporal alignment between videos, but also the preferences and social relationships of each individual user. This subject is the focus of the next chapter, which discusses the
Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance? Figure 4.1. Initial prototype implementation for automatic video editing.
Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance? Figure 4.1. Initial prototype implementation for automatic video editing.
Figure 4.2. Initial prototype implementation for manually editing videos.
Figure 4.2. Initial prototype implementation for manually editing videos.
Figure 4.3. Elements of an authored video composition: the parent has included an introductory image and a video for making it more personal and intimate.
Figure 4.3. Elements of an authored video composition: the parent has included an introductory image and a video for making it more personal and intimate.
Based on participants’ comments, reactions, and answers to the questionnaires, we can conclude that they appreciated the benefits of our authoring system and considered it a valuable vehicle for creating enjoyable memories. While these results were highly relevant, we were aware that they were not complete. More importantly was the indication that instead of the automatic or the manual authoring thread, a hybrid solution would better fit the participants’ needs (Q1.4). In the next section, we discuss the functional requirements that motivated the design of a new version of our socially-aware multimedia authoring system.
Based on participants’ comments, reactions, and answers to the questionnaires, we can conclude that they appreciated the benefits of our authoring system and considered it a valuable vehicle for creating enjoyable memories. While these results were highly relevant, we were aware that they were not complete. More importantly was the indication that instead of the automatic or the manual authoring thread, a hybrid solution would better fit the participants’ needs (Q1.4). In the next section, we discuss the functional requirements that motivated the design of a new version of our socially-aware multimedia authoring system.
Figure 4.6. High-level workflow of our hybrid authoring tool.
Figure 4.6. High-level workflow of our hybrid authoring tool.
Figure 4.7. Triggering the Director’s Cut from the browsing interface. Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance?
Figure 4.7. Triggering the Director’s Cut from the browsing interface. Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance?
Figure 4.8. Comparison between the recording behavior of a mother towards her kid (in blue) and the average behavior of the rest of the parents (in red). Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance?
Figure 4.8. Comparison between the recording behavior of a mother towards her kid (in blue) and the average behavior of the rest of the parents (in red). Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance?
Figure 4.9. Director’s Cut: an initial video compilation is created automatically by the system.
Figure 4.9. Director’s Cut: an initial video compilation is created automatically by the system.
Figure 4.10. Director’s Cut: visualizing alternative clips available. Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance?
Figure 4.10. Director’s Cut: visualizing alternative clips available. Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance?
Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance? All participants indicated they would like to create different types of productions (Q2.7). When questioned about the types of video stories they envisaged, the ‘song-based’ video came up as first choice among most of them. Some argued that depending on the social situation they would create and share different versions with different audiences.
Chapter 4. Automatic and Manual Processes for Creating Personalized Stories from Community Assets: Where is the Balance? All participants indicated they would like to create different types of productions (Q2.7). When questioned about the types of video stories they envisaged, the ‘song-based’ video came up as first choice among most of them. Some argued that depending on the social situation they would create and share different versions with different audiences.
Figure 5.1. Typical end-user comment in YouTube. It appears statically underneath the third-party video.
Figure 5.1. Typical end-user comment in YouTube. It appears statically underneath the third-party video.
Figure 5.3. Requirements gathering: utility and usefulness of timed comments within media. Media Consumption and Commenting Practices
Figure 5.3. Requirements gathering: utility and usefulness of timed comments within media. Media Consumption and Commenting Practices
Figure 5.5. Timed text content and temporal hyperlinks.
Figure 5.5. Timed text content and temporal hyperlinks.
A Timed Text Video Commenting System
A Timed Text Video Commenting System
Chapter 5. Supporting Personalized End-User Comments within Third-Party Online Videos HIpul dled UIisure 9.6 (O)). One key feature of our video commenting system is its ability to (semi-) automatically compute the temporal alignment of user-generated comments. To explain how this functionality works, consider the example in Figure 5.8. In a usage scenario, we assume users will interact after a certain moment of interest has passed (e.g., after seeing the trombone on screen). In this case, comments need to be synchronized in such a way as to avoid situations in which the comment - “Look at that shiny, beautiful trombone!” — appears after the instrument is no onger visible. Our approach for this use case is as follows. When an end-user indicates s/he wants to add a comment, the video playback is paused and the inpu area gains focus (Figure 5.8 (2)). As the interaction is performed right after istening to or watching an event of interest, we assume that the current momen (t,o) is the end of the comment (t.yg = thw). AS an initial guess, we consider that the start time of the comment (t,..,,) is equal to the current time (t,,,,) minus a minima duration (MinDur) that a comment should stay on screen for being effectively read (tsart = touess = thow - MinDur).
Chapter 5. Supporting Personalized End-User Comments within Third-Party Online Videos HIpul dled UIisure 9.6 (O)). One key feature of our video commenting system is its ability to (semi-) automatically compute the temporal alignment of user-generated comments. To explain how this functionality works, consider the example in Figure 5.8. In a usage scenario, we assume users will interact after a certain moment of interest has passed (e.g., after seeing the trombone on screen). In this case, comments need to be synchronized in such a way as to avoid situations in which the comment - “Look at that shiny, beautiful trombone!” — appears after the instrument is no onger visible. Our approach for this use case is as follows. When an end-user indicates s/he wants to add a comment, the video playback is paused and the inpu area gains focus (Figure 5.8 (2)). As the interaction is performed right after istening to or watching an event of interest, we assume that the current momen (t,o) is the end of the comment (t.yg = thw). AS an initial guess, we consider that the start time of the comment (t,..,,) is equal to the current time (t,,,,) minus a minima duration (MinDur) that a comment should stay on screen for being effectively read (tsart = touess = thow - MinDur).
Figure 5.9. Predictive timing support for end-user comments.
Figure 5.9. Predictive timing support for end-user comments.
Figure 5.11. Results from the close captioning activity using our prototype system.
Figure 5.11. Results from the close captioning activity using our prototype system.
Related topics:
User-Generated ContentMultimedia AuthoringSocial connectednessUser Generated ContentSynchronized Multimedia Integration Language
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved