-
00 : 01
© Distribution of this video is restricted by its owner
Transcript
×
Auto highlight
Font-size
| 00:03 | Hello. My name is just balsa and this is a short video to |
|
|
|
| 00:09 | you an idea about my research. research is in the context of a |
|
|
|
| 00:15 | called Video points First. I'll get why we decided to go into this |
|
|
|
| 00:24 | of research as you all know. is a growing as a technology to |
|
|
|
| 00:31 | and sometimes replace live lecture. So is extremely important, especially now as |
|
|
|
| 00:37 | instructional tool or an instructional medium. the most important shortcoming of the video |
|
|
|
| 00:44 | is that it's extremely difficult to quickly to the content that interests you the |
|
|
|
| 00:51 | to your question in a video. we've done a lot of work on |
|
|
|
| 00:57 | various uh tools and methods to make essentially what I call an interactive learning |
|
|
|
| 01:10 | . So now I'm going to show few screenshots from this system. You |
|
|
|
| 01:15 | have see this demo of the system at this that site called Video points |
|
|
|
| 01:22 | org. And this is something that being used quite a bit at the |
|
|
|
| 01:27 | of Houston. So one of the interesting things Video points does is called |
|
|
|
| 01:33 | . So here you see the length the timeline of of a lecture video |
|
|
|
| 01:39 | is an hour plus and it's been up into this subsection and each one |
|
|
|
| 01:46 | um um a subtopic within that And for each of these segments there |
|
|
|
| 01:55 | a frame called the summary frame. is currently is highlighting the summary frame |
|
|
|
| 02:01 | believe for this segment, but it be another one and the summary frame |
|
|
|
| 02:05 | of the keywords that are the most for that video as well as a |
|
|
|
| 02:11 | of representative images? So that's segmentation summarization. Another interesting feature of the |
|
|
|
| 02:21 | is multi model search where we can a search term here. The system |
|
|
|
| 02:29 | show all these parts of the video the search has a match. The |
|
|
|
| 02:35 | of matches is captured by this white green bar and the matches are also |
|
|
|
| 02:41 | on the transcript of the audio for video and here are some examples of |
|
|
|
| 02:50 | . There is also support for This uh image is showing this usage |
|
|
|
| 02:57 | in terms of the number of duration , number of accesses. So it's |
|
|
|
| 03:02 | for an instructor to track how students using a lecture video. I'm going |
|
|
|
| 03:09 | give one or two example research challenges terms of what goes under the hood |
|
|
|
| 03:16 | make all of this possible. So central challenge is how to automatically divide |
|
|
|
| 03:21 | lecture video into segments where each segment a subtopic. We could look at |
|
|
|
| 03:28 | if the a section of video has images, it is likely to be |
|
|
|
| 03:35 | the same or similar topics. we could look at the text on |
|
|
|
| 03:39 | screen or speech and see the patterns the change of vocabulary in there. |
|
|
|
| 03:45 | might give us an idea of where topics are the same and where our |
|
|
|
| 03:50 | changes. So a related problem is . Now I have a video |
|
|
|
| 03:56 | I want a single frame summary of segment. So one part of the |
|
|
|
| 04:02 | could be key words. But then do I find out these keywords? |
|
|
|
| 04:05 | can look at the frequency of I can look at feel important. |
|
|
|
| 04:10 | words for example, like python or have a special meaning in computer science |
|
|
|
| 04:16 | if that's a factor and I could look at on size, location on |
|
|
|
| 04:20 | and whatnot and the challenges to how we put all these together to get |
|
|
|
| 04:24 | best keywords? I also want to the best images to represent a lecture |
|
|
|
| 04:31 | segment and there again, what constitutes best image? I could look at |
|
|
|
| 04:37 | size time on the screen complexity and on as well as uh, we |
|
|
|
| 04:43 | to look at the similarity between images a good summary is not repetitive. |
|
|
|
| 04:48 | image is unique and represent something. this gives you an idea at a |
|
|
|
| 04:53 | level of the challenges and the underlying . We have to do tax |
|
|
|
| 04:59 | speech analysis, image analysis to work user interfaces as well as do service |
|
|
|
| 05:06 | see the real world impact of this research. And this also involves by |
|
|
|
| 05:12 | way, what's commonly known as artificial and machine learning, which is a |
|
|
|
| 05:18 | of analysis we are talking about. we have projects for all levels of |
|
|
|
| 05:24 | as I mentioned earlier, you can out more about the system at video |
|
|
|
| 05:29 | dot org and for everything else, contact me and we can have a |
|
|
|
| 05:35 | and see if this is an area research that is of interest to |
|
|
|
| 05:41 | So that concludes this |
|
|
|
A simple tooltip for the link