Assignment 3: Video Production for Research Videos

by Mike Gleicher on April 1, 2011 · 59 comments

in Assignments

Assignment due on Thursday, April 7th – so we can discuss it in class on Friday, April 8th.

Video is a big part of dissemination of research results. Not just in graphics, but in HCI, multimedia, visualization, computer vision, …

Video is important to accompany a paper: we often submit videos to accompany submissions, and afterwards they are useful for people to see how the results mesh with the paper. Often, people will look at the video first, so making a good impression and getting the main ideas across is critical.

There are also “video papers” – videos designed to stand alone, introduce the basic ideas, show off the results – possibly independently of the paper (or, often as an advertisement to go read the paper). These are becoming increasingly important. Far more people will watch your video on YouTube than read your SIGGRAPH paper.

Videos used to be a niche thing, but now that it is so easy to spread them (via YouTube and the like), they are really common. Back in the old days you used to have to mail VHS video tapes (which had poor quality as well as taking up shelf space and mailing costs).

The thing that hasn’t changed is that its hard to produce good research videos. Yes, the technical challenges are gone with modern screen capture software, you’ll never have to figure out how to get a scan converter to connect to your VCR, you can do non-linear editing and see real-time previews on just about any computer, the software is easy to use, … (editing video with VCRs is not easy).

Making a video is like writing: just because you can make a sentence doesn’t mean you can tell a story or communicate your point.

We do a lot of paper reading, and talk about paper writing. But we don’t do enough with videos. Fortunately, videos are short, so you can watch several of them.

As a first step into this, I want everyone to watch a bunch of research videos and think about the presentation.

The Assignment

Your assignment is to watch at least 4 different research videos (and another one below). Two must come from my list below, and two others you need to find yourself (subject to the rules below). For each one, you must post a comment to this page (so you’ll make 4 comments) answering the questions I’ll list below. Some you’ll need to write more than others.  The assignment is Due Thursday, April 7th (any time before Friday, April 8th). Yes, I know the project is going on too – but you haven’t had much reading to do. And this exercise will help with the project (and, in fact, get you some extra time to use what you’ve learned for your project).

I will also give you a “bad” video (it’s one that I produced hastily). I want you to watch it, and think about why it doesn’t work as well as the well-produced videos shown before. It will be particularly hard since its an accompaniment to a paper you haven’t read (well, Danielle wrote it and some of you helped proofread it). I prefer to give you my own stuff for the bad examples as not to embarrass someone else.
Here is a link to a video I made for a recent paper submission. Danielle can tell you about the actual work being depicted. But for now, view it as an example of a video where the creator (me, not Danielle) did not put the time and effort required to make something good. To be honest, this quality level is about the level of expectation I have for the project one videos that you need to make.

Here are some suggestions of videos to watch (watch at least 2):

  • Content-Preserving Warps for 3D Video Stabilization. (project page, YouTube) – yes, this is my work, but the video was made by Aseem Agarwala, who is quite good at making these videos.
  • Animated Transitions in Statistical Data Graphics. (project page). A well done video about some interesting visualization research.
  • Stable Spaces for Real-time Clothing
  • Parametric Motion Graphs. (Project page with downloadable video, Youtube – our video, but uploaded by someone else) – This was produced by my former student, Rachel Heck, who in addition to doing the great animation work, was really good at presenting it). The video is more of a “results video” (to accompany the paper) than a video paper that introduces the technique.
  • Retargeting Motion to New Characters – one of mine. (Youtube, AVI download) – One of my old ones. Old, but I think it was pretty effective at telling the story.
  • Spatial Relationship Preserving Character Motion Adaptation (SIGGRAPH 2010) (youtube)
  • Rigid Body Fracture Sound (YouTube) – interesting because the “results” are sounds.

This video is only showing the data used in a study (although its interesting). You can learn something comparing this to the more “video paper” like things.

Places to look for videos (please pick videos created by researchers to share their results – preferably connected to a paper, or meant to be a “video paper”):

  • Look at the project web pages for papers you have read. Quite often, you’ll find a video. In fact, I suggest watching the video for at least one paper you’ve read recently. Many of the papers you’re using for projects have good videos associated with them.
  • Look at the acm digital library page for newer SIGGRAPH papers. Usually, you can watch the video of the talk. However, under “source materials” you can find the videos.
  • Our group’s video page has lots of old videos. Many of them were pretty well done (the technical hassles made it so much work to produce any video, you might as well have made a good one).

Here are some questions I want you to think about:

  1. Be clear about what the video is (name, type, link to where you found it on the web (preferably a page with the video, not just a link to the video itself).
  2. In a sentence or two, explain what the video tells the viewer.
  3. Your overall impression of the video – how well did it make its point? How effective is it at conveying its ideas?
  4. What kinds of things about the work did it convey?
  5. Would it stand by itself, augment a paper, serve as an advertisement for a paper? Why?
  6. What did you notice about the video itself (in terms of its production)? Anything from quality of the audio, legibility, editing, pacing,  … Just like with writing, a lot of what makes video good can be very subtle. Try to notice what kinds of things influence you.

For all of the videos you watch (at least 4), please answer questions 1-3. For at least 2 of the videos, answer all 6 questions.

We’ll talk about film and video production in general (editing, cinematography, …) later in class. For now, I want everyone to just get a sense of what a research video is.

{ 57 comments }

xlzhang April 5, 2011 at 4:23 pm

1. Content-Preserving Warps for 3D Video Stabilization; SIGGRAPH 2009;
http://pages.cs.wisc.edu/~fliu/project/3dstab.htm

2. This video talks about existing methods for stabilizing videos; describes
the new proposed method; contrasts the results of these various methods.

3. The video was very effective in showing that the proposed method is
better than existing methods for stabilizing video in various situations.
It also honestly described the method’s shortcomings when applied to
video with a spiraling movement. In terms of conveying its ideas, the
narration was a bit fast when describing the technical details, but the
video did a great job of eliciting interest in the paper itself.

4. The video successfully conveyed that the problem being addressed is
important not just to researchers, but to a commercial market interested
in significantly improved video stabilization for home videos; who knows,
maybe even low budget film producers.

5. The video is best used as an advertisement for a paper because of its
level of detail. It focused mostly on demonstrations of how much better
the new method is than existing methods, rather than detailed
descriptions of the actual algorithms or code. It also effectively
augments the paper, because the effects of the paper are difficult to
convey using static photos.

6. The video had very professional sound editing, as opposed to, say, the
video on Spatial Relationship Preserving Character Motion Adaptation.
Clearly, a lot of effort was put into the after effects, which also made
it seem more professional.

xlzhang April 5, 2011 at 4:24 pm

1. Stable Spaces for Real-Time Clothing; SIGGRAPH 2010;
http://drp.disneyresearch.com/projects/cloth/

2. This video describes the researcher’s proposal for a real-time approach
to realistic cloth animation.

3. There is only one sentence at the beginning that tells the viewer
any details about their approach, so the viewer may start out confused
as to what the differences are between the new approach and other
methods. Other than that the demonstrations are very effective.

4. The video conveys that the work is focused on improving the performance,
speed, efficiency of clothing animations. When compared side by side
with results from Maya it is not dramatically better, but the clip at the
end, which shows a moving army of clothing, effectively conveys how fast
the method is.

5. This video could not stand by itself due to its lack of detail about
the method, however it is good for augmenting the paper because the
results can be rather subtle unless viewed in motion.

6. The pacing could have been better. One thing I liked about the Content
Preserving Warps video was that there was rarely a lull in the narration,
so that it felt as though the researchers had a lot to say, and the
timing was more streamlined. Perhaps out of necessity, the examples were
very repetitive, and the video may have been better if the motions were
more varied.

xlzhang April 5, 2011 at 4:24 pm

1. Video-Based Reconstruction of Animatable Human Characters; SIGGRAPH 2010
(Asia); http://www.mpi-inf.mpg.de/resources/perfcap/index_vrhc.html

2. The video describes an approach to create a fully animatable rigged
model of a person. The model is generated from multi-view recordings,
and seems to generate something akin to a bvh skeleton, skinned to look
like the model, and even correctly incorporates the clothing they are
wearing. No markers are used in the video capture.

3. For me this video raised more questions than it answered, but the topic
is an interesting one. For instance, the system can successfully capture
the subject, segment the cloth bits away from the body bits, extract a
skeletal rig based on the visible body bits. The skeleton rig is then
manipulated to produce new animations, and the skin is reapplied so that
the result looks like the human model performing the new motion. At each
stage I must ask “How?” and so this video is definitely mostly an
advertisement for the paper.

xlzhang April 5, 2011 at 4:25 pm

1. Simulating virtual environments within virtual environments as the basis
for a psychophysics of presence; SIGGRAPH 2010;
http://graphics.cs.wisc.edu/WP/virtualreality11/2011/03/01/simulating-virtual-environments-within-virtual-environments-as-the-basis-for-a-psychophysics-of-presence/
I could only find the video here, it doesn’t seem to be linked on the
author’s homepage.

2. I have no idea. It starts with the title, what does “a psychophysics”
mean? I can look up the definition of psychophysics but not “a”
psychophysics… It seems to just show the process of their experiment:
insert guy in virtual environment, turn on / off various signals to
the simulator so that he either can or can’t seem himself moving, or
see himself period. The guy spends most of the time looking confused.
Perhaps they do something with this data later but I guess that’s hard
to show.

3. I probably just didn’t understand what they were trying to show with the
video, but I could not determine the purpose of the *video*. I can
somewhat understand the purpose of the research by reading Mel Slater’s
website and his interests, but not the purpose of the video, unless it
was simply to show their methodology for collecting observations and data.

Aaron Bartholomew April 5, 2011 at 7:34 pm

1) Rigid-Body Fracture Sound (http://www.youtube.com/user/DougLJames#p/u/0/nHH8N_lNZzI)

2) The video presents examples of results from the authors’ rigid body fracture + sound algorithm using different parameter inputs.

3) This video is not very effective at conveying its ideas, especially for the sound parameter results. Although the fracturing looks great, the authors give no inclination as to why it is good (a greater quantity of great-looking examples still doesn’t clarify anything). Surprisingly, the explanation of the algorithm’s sound capabilities is even worse. They present a sequential stream of examples with different parameters, but the differences are sometimes so subtle that it would seem nothing has changed. The authors need to explain what is happening and why it is better.

4) That the algorithm can simulate fracturing of rigid bodies with accurate collision sounds in real-time. The collision sounds are correctly updated for different surface materials and observer distance.

5) This video would serve best as an addition to a paper that explains the algorithm (maybe the paper goes in depth about specific examples in the video) or maybe as an advertisement to show how cool it is (but should definitely be shortened/not go into details).

6) In terms of editing, I think the sequence of events is logical: start with examples of what the system can do, then gradually go into greater detail for why they are possible. Just like with research papers, by presenting an abstract of what the system does, it will be easier to garner interest and quickly summarize the solution to those who don’t want to watch the whole thing. Finally, it seems like the author believes in quantity over quality; there are far too many rushed examples. It would be more effective if the author devoted time to fewer examples with in-depth discussion of major points. Thus, with some author narration/explanation and less examples, this could be a good video.

Aaron Bartholomew April 5, 2011 at 7:35 pm

1) Parametric Motion Graphs (http://pages.cs.wisc.edu/~heckr/Research/pmg.html)

2) This video first explains the general idea of what parametric motion graphs are (by showing random walks of different motion sets and adaptation to target constraints), and then demonstrates how this system can be used for interactive control.

3) I found this video to be highly effective in conveying its ideas. The flow of information was perfect in the sense that it naturally progresses from underlying concepts to the complex capabilities of the system. Rachel’s presentation of random walks and target constraints establish the notion of connections and graphs searching such that they serve as a foundation for inferring the interactive graphs later in the video. Thus, even someone unfamiliar with motion graphs can formulate a sense of why this works. Also, the inclusion of someone controlling the character with a joystick sums it up nicely with the project’s ‘end product’.

Aaron Bartholomew April 5, 2011 at 7:35 pm

1) 2.5D Cartoons (http://www.alecrivers.com/2.5dcartoonmodels/)

2) This video explains: how 2.5D cartoon models work, how to make them using their system, and what their limitations are.

3) I’d say this video is pretty much as good as it gets. Maybe it’s by simplicity of the method’s design, but the authors were able to provide some visuals that really make the underlying concept clear (such as showing how the billboards update with camera rotation). In addition, the narration is concise (explaining only key points) which is suitable since the visuals are good enough to do the talking.

4) Pretty much everything that needs to be known (how it works, how to make it work, and when it doesn’t work).

5) This video can definitely stand-alone. Again, it’s probably because the underlying concepts are so simple, but the authors were able to effectively communicate everything needed for comprehension (I feel like I don’t even have to look at the paper now).

6) The portion of the video that brought everything together for me was the inclusion of the model creation process (a great pacing decision). This video was like a narrative, with the creation process as the climax and the system limitations as the resolution; the underlying concepts were all brought together in a simple to follow, wrap-up demonstration. Also, I felt like this video was highly ‘efficient’ in conveying information. Examples were shown just long enough to demonstrate a key point and there were enough unique ones to keep things interesting.

Aaron Bartholomew April 5, 2011 at 7:36 pm

1) Skinning Mesh Animations (http://graphics.cs.cmu.edu/projects/sma/)

2) This video tells the viewer how their method computes weights for skinning a mesh animation and provides examples to motivate its use (collision detection and graphics hardware computation).

3) This video was decent, but at times I question if I had my AVI player set to 1.5x speed. The authors certainly don’t waste time; they immediately get into the details of their algorithm for weight calculation without ever really explaining what they are for. Their purpose comes from the viewer’s inference rather than a concrete explanation from the author, leaving some residual ambiguity. Other than that though, the explanation of the algorithm itself if very clear and serves to gradually clarify the method’s purpose.

Michael Correll April 6, 2011 at 11:34 am

1. Animated Transitions in Statistical Data Graphics, from the given link. http://vis.berkeley.edu/papers/animated_transitions/

2. The video illustrates different techniques for viewing changes in the perspective or content of statistical data using animation, as well as presents the experimental results from users interacting with these systems.

3. I thought the video was very well put together, and presented a clear narrative. One of the gaps in the video was the motivation and the theoretical background of the work: it’s very easy to get the idea that they just tried a couple things and then see if it worked out, instead of rigorously investigating what works and does not in using animation in information visualization.

4. It conveyed the power of the animations, the performance and feedback gains of animated transitions, and the depth of the space of constructing animated transitions.

5. By the nature of the topic of the research, a static paper would do a poor job at conveying the ideas presented. (Ironically, if a static paper did an equally competent job of presenting the results of their work, it would provide strong evidence that their work was misguided). While it certainly augments the work, I would want to see a lot of motivating research and related works before passing judgment on the work as a whole, items which don’t really belong in a video.

6. I noticed that the video did a good job of letting the animations speak for themselves: the narration almost never “collided” with novel transitions. I also like that there were clear sections in the video (introduction and overview, staged vs. single animations, results) and yet there were not explicit title slides separating the sections: the flow of the narrative dictated the section transitions rather than trying to shoehorn them in abruptly.

Michael Correll April 6, 2011 at 11:50 am

1. Parametric Motion Graphs, again from the given list,
http://www.youtube.com/watch?v=ZepNQHXAcCY

2. The video does a good job of saying “this technique we developed generates some compelling results, in real time.”

3. I think the video is compelling as a sales pitch, whetting the appetite for attending the associated talk and/or reading the associated paper. I don’t know if it does a good job of conveying ideas, especially since it is so focused on the results rather than the process. There’s no idea of how hard this problem is to solve (although in the context of the conference venue I’m sure the benefits and comparative quality of the results are known).

4. The video conveys that the parametric graph technique is flexible, interactive, robust, and real-time.

5. As with the animated transitions video (above) the nature of the technique and the generated results means that a video is inherently a better vehicle for displaying results than a static paper. However, where the transitions video presented an overview of both the techniques and the results, this video did not go in depth into the actual technique (which, granted, is difficult to visualize in a time and space-compact way). It’s definitely more of a sales-pitch video than an overview video.

6. There was definitely a building block structure to the video, starting with basic results and using those as a grammar on which to build to more complex results. The use of title slides to show the different sorts of results generated (and how they differ in both type and complexity) means that even without sound it is possible to get most of the ideas out of the video. In fact the sound is almost redundant and doesn’t contribute much beyond brief orientation.

Jim Hill April 6, 2011 at 1:16 pm

1. Content-Preserving Warps for 3D Video Stabilisation (http://pages.cs.wisc.edu/~fliu/project/3dstab.htm)
2. This video demonstrates a method for stabilising video shot with non professional recording equipment. The method attempts to reconstruct the path of the camera and warps the image to reduce camera jitter.
3. The video showed some very impressive results. It made it’s point well and seems to be a very good method for quickly and cheaply stabilising video.
4. Mostly this video conveyed the results of the work. It also gave a brief explanation about how the method worked computationally and some of the drawbacks to the method.
5. I don’t believe that it could completely stand on it’s own. At least, someone would have to figure out a lot of math before implementing it if they didn’t have the paper. It would certainly augment the paper.
6. I liked that the speaker was clear and didn’t talk too fast. I also liked that he didn’t try to use buzz words or jargon, it made the video much easier to watch. The visual methods for demonstrating the method where also very nicely coordinated.

Jim Hill April 6, 2011 at 1:16 pm

1. Stable Spaces for Real-Time Clothing (http://drp.disneyresearch.com/projects/cloth/)
2. The video is basically an advertisement for real-time clothing simulation. It presents the results of the fast clothing simulation but not much else.
3. The results of this video were impressive, but it did almost nothing to actually explain the method. There was a blip in the beginning but it didn’t make much sense and created more questions than answers.
4. It really only conveyed that the results looked good and where generated quickly.
5. This video could not stand by itself and it wouldn’t really augment the paper either. This is definitely an advertisement.
6. The message was clear, although a lot of jargon was used in the beginning. The demo’s where nice and showcased the results nicely.

Jim Hill April 6, 2011 at 1:16 pm

1. Synchronised Multi-Character Motion Editing (http://mrl.snu.ac.kr/research/ProjectMultiMotionEditing/index.html)
2. This video shows off a system built to produce synchronised character animation.
3. This is a very complicated video that attempts to show off a lot of stuff. It’s makes it’s point very clearly and effectively, the results are very impressive as is the interface created to get those results.
4. It basically tried to show all of the things that their system could do. It didn’t really focus on how it all worked.
5. It definitely augments the paper, I don’t think you could get enough from it to recreate the paper though.
6. The speaker talked slowly which made it hard to listen to. Other than that, the interface was attractive and the demonstrations where easy to follow.

Jim Hill April 6, 2011 at 1:17 pm

1. Virtual Videography (http://www.cs.wisc.edu/graphics/Gallery/VirtualVideography/)
2. This video tells the user about the generation of edited video of lectures.
3. The video definitely conveyed the interesting ideas. It was well directed.
4. It basically conveyed the results of the work without going into the gory details.
5. This seems like an advertisement for the paper. There really wasn’t much detail on the actual implementation, just what the system did.
6. I thought the speaker had a good tempo and explained things clearly. The methods used to show the different camera views was also very nice.

Michael Correll April 6, 2011 at 2:08 pm

1. I saw the Scatterdice video, http://www.youtube.com/watch?v=2bYIRcO-gwg&feature=related
based on the paper Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation, available http://www.lri.fr/~dragice/papers/scatterdice.pdf

2. The video communicates the details of a system that allows navigation and visualization of high dimensional scattered data, with filtering, querying, and detail on demand abilities. It also illustrates a use case where a user weighs different data dimensions (cost, zoom, resolution, &c.) to choose a camera.

3. The use case I feel was a little too bare bones: it was hard to get the motivation behind the effort until the very end. In particular, since it came after the in depth description of features, it didn’t present anything novel beyond seeing an example of high dimensional data. I think it would’ve been better to start out with a use case, and then progress from that use case to a list of features, describing features as they organically arise from the user’s actions.

Michael Correll April 6, 2011 at 2:46 pm

1. I also watched the video associated with the 2.5 Cartoon Paper that was an (optional) reading from earlier this semester: http://www.alecrivers.com/2.5dcartoonmodels/

2. This video illustrates how interpolation strategies can be used to generate quasi-3D models based on stroke-based drawings from relatively few viewpoints.

3. The video I found a lot more compelling than the paper per se (the paper left me feeling “is that it?” which is usually not a good sign). The methods being used are not detailed very well but as an “advertisement” video I don’t really expect them to be. Providing the interactive rotating models generated by the system was a lot more compelling then just showing some 2D stroke interpolation in the paper (even though that’s really the core of the technique). The video was also narratively “flat,” that is, rather than showing simple cases and building to more complex use cases, it really just showed a series of similar examples in sequence. Again, that’s okay for a paper-ad style of video, but it makes the technique seem shallow and non-extensible.

sandrist April 7, 2011 at 12:47 am

1. Content-Preserving Warps for 3D Video Stabilization; SIGGRAPH 2009;
http://pages.cs.wisc.edu/~fliu/project/3dstab.htm

2. This video presents a novel technique for stabilizing home-made videos so that they can look more professional. It compares this technique to both a baseline (no stabilization) and the current state-of-the-art, with favorable results.

3. I had a very good overall impression of the video. It was enjoyable to watch and made its points clearly and succinctly. I appreciated the final minute or so of video comparisons set to music, since I think most video makers wrongly feel the need to speak as fast as they can for the entire 5 minutes and tell you everything they possibly can.

4. This video focused mostly on the results of applying their new technique, but spent a small amount of time giving the intuition behind it. With content-preserving warps created from minute changes in viewpoint at each frame, individual frames might look a little funky, but the final stabilized video looks quite nice.

5. Since the video didn’t go into very many details at all, and focused instead on showing off the nice shiny results, I would say this video is a very effective advertisement for the paper.

6. I thought the video production quality was overall pretty good. He might have spoken a little quickly in the beginning, but I appreciate the fact that he let the results speak for themselves by the end.

sandrist April 7, 2011 at 1:01 am

1. Rigid-Body Fracture Sound (http://www.youtube.com/user/DougLJames#p/u/0/nHH8N_lNZzI)

2. This video tells the viewer that these researchers have come up with a way to make things that are breaking sound more like real-life things that are breaking, in real time.

3. This video was definitely not very effective from the point of view of someone going straight to the video and not knowing anything else. I saw things breaking for 3.5 minutes with labels on the side displaying which parameters were being tweaked or which techniques were being compared. None of these parameters or techniques had any explanation, so they were completely meaningless to me.

4. This video focused completely on results, showing the same 5 or 6 things being broken over and over again. It was unclear to me exactly what was going on most of the time.

5. This video could certainly augment a paper, as long as it is made very clear that this video makes no sense without having at least skimmed the paper or read its abstract first. It wasn’t very effective as an advertisement for me, since I couldn’t glean anything compelling from it that made me excited to learn more. Any attempt at a narrative would have helped.

6. The video production quality was good, considering how simple it was. Putting blocks of text next to brief animations is not effective, since I can’t simultaneously watch the animation and read its explanation, but I need both to know what’s going on. Again, any sort of narration to convey a sort of storyline would have helped a lot.

sandrist April 7, 2011 at 1:09 am

1. Motion Fields for Interactive Character Animation
http://grail.cs.washington.edu/projects/motion-fields/

2. This video presents motion fields, and shows the viewer how much more responsive to control it is than previous techniques, while still producing nice looking animation.

3. I like this video a lot, since I think it highlights very well the main improvements that motion fields offer over other techniques such as motion graphs. The animations are shiny and smooth, the narration is well-paced and understandable, and the results shown are compelling.

sandrist April 7, 2011 at 1:16 am

1. As-Rigid-As-Possible Shape Manipulation
http://www.youtube.com/watch?v=1M_oyUEOHK8

2. This video walks the viewer through the steps involved in the researchers’ novel technique for shape manipulation, and present some nice results as well as comparisons to previous techniques.

3. I think this video is effective at showing off the as-rigid-as-possible technique, as well as giving the key insights into how it works without going into a lot of detail. They make some weird comparisons to naive versions of their own technique, but overall I think they make a good argument for the usefulness and effectiveness of their system.

danieljc April 7, 2011 at 1:26 am

1. Parametric Motion Graph (http://www.youtube.com/watch?v=ZepNQHXAcCY)

2. The video demonstrates the quality of interactive results based on parametric motion graphs. It consists of a series of demos and focuses on showing the results rather than the behind the scenes technical aspects.

3. It made parametric motion graphs look like a really useful idea that provides good results, so I would say it was quite successful. It is very clear what is happening throughout the video.

4. The smooth transitions from one motion to another, and the ability to fit the motions to a specified pattern seem to be the major ideas conveyed.

5. It stands by itself in the sense that a viewer could understand the user of parametric motion graphs without reading the paper. At the same time, the video wouldn’t really help that much in understanding how parametric motion graphs actually work as it doesn’t deal with the technical discussions on a detailed level. So it seems generally more like an advertisement

6. In general the pacing was quite good and the video looked professional. The ending shot with the joystick was a little frustrating, since the quality of the shot didn’t quite match the direct capture of the rest of the video and the motion results didn’t seem as “good” in that shot as the rest of the video, probably a result of lag that gets hidden when you are only seeing the video without the controller onscreen.

danieljc April 7, 2011 at 1:28 am

1. Retargeting Motions to New Characters (http://www.youtube.com/watch?v=Vn-vVzMGgec)

2. The video shows how a motion can be retargeted to a new character. Several different changes, including changing sizes of characters and using animation from an entirely different type of object are demonstrated.

3. This video quickly got across the uses of the techniques it was showing. I was surprised how the video spent so much time showing potential bad results. In a sense, it could almost make me trust the rest of the results more since it seems like the drawbacks are being so clearly shown.

4. It conveyed the flexible ways in which character animation can be applied to different sized or shaped objects, even objects that change as the motion is played. It also included some technical discussion about the techniques, though that did not seem to be the major focus of the video.

5. This video seems like it augments the paper, although portions also seem like an advertisement. If the video was solely for an advertisement, it would likely have included fewer drawbacks that can occur and less technical content.

6. I noticed how two or three angles were used multiple times throughout the video. One minor thing that bothered me is when it was said that “…when the character is changing size…” – I couldn’t tell clearly that the size was actually changing in the video, or if it wasn’t matching up with the audio.

danieljc April 7, 2011 at 1:34 am

1. Responsive Characters from Motion Fragments (http://graphics.cs.cmu.edu/projects/rcfmf/)

2. The video attempts to show the results of the various techniques discussed in the paper, including the improved responsive results from the motion fragments compared to older techniques. It also contrasts several learning algorithms.

3. My impressions of the video were not especially positive. First of all, there was no audio (that I could hear at least), which sometimes made it hard to fully understand what they were trying to show. The general setup involved showing what different types of learning algorithms would produce. The environment shown in the video made it a little difficult to evaluate either, with characters skating up and down, non-animated, over height changes.

danieljc April 7, 2011 at 1:37 am

1. Skinning Mesh Animations (http://graphics.cs.cmu.edu/projects/sma/)

2. This video shows a use of an automatic generation of mesh animation using linear blend skinning. The video is more focused on the technical aspects of the paper.

3. There are some interesting clips within it showing some good results from the technique, but it is more technical based, with rapid explanations throughout the video that could be a little hard to follow at times. This video really requires reading the paper to fully understand the video, but it is still quite possible to understand visually many of the results they are showing without that knowledge.

sgallege April 7, 2011 at 7:59 pm

1. Parametric Motion Graphs, http://pages.cs.wisc.edu/~heckr/Research/pmg.html
2. Results of using parametric motion graphs to drive a character interactively
3. I think it is a good video over all, and conveys the point well. There are some aspect s that could be improved. It provides a variety of example with narrative to make it’s point.
4. It conveys how the new technique can be applied to various animation sequences including random, interactive, and different kinds of motions by specifying direction, speed, target etc.. . It also stresses that the motions are smooth and seamless
5. I think this would mainly serve as a advertisement and also augment a paper. It’s not a standalone video as it doesn’t provide any implementation details
6. The visuals, narrative and captions are well synchronized, and the clarity of visuals and narrative are very good. The back ground music was somewhat distracting at times as it was interfering with the narrative. The transitions could have been better than a black out, a title slide would have been nice. Due to the lack of comparison to a different technique the viewer cannot evaluate how good the parametric motions are.

sgallege April 7, 2011 at 8:00 pm

1. Stable Spaces for Real-time Clothing http://drp.disneyresearch.com/projects/cloth/
2. A technique for learning clothing models that enables the simultaneous animation of thousands of detailed garments in real-time
3. I think this a very good video overall. It provides a very good introduction, many examples and comparisons with different parameters and to existing techniques
4. It conveys the ability render real time cloth animation with dynamic effects various dimensions and materials. They also provide an introduction to the technique used in the paper. Finally they display the efficiency of the technique by showing a real time rendering of multiple simulations.
5. This can be considered a standalone video as it contains an introduction to the technique used in the paper and comparisons with various parameters.
6. The visuals, narrative and captions are well synchronized. The section titles helpful in making smooth transitions. The narrative is well paced and very clear. The side by side comparison work very well to convey the improvements of the new technique, also the narrator makes sure to draw attention to the important parts of the comparison. I think this video has got all the elements to work correctly.

sgallege April 7, 2011 at 8:00 pm

1. Example-Based Facial Rigging http://www.hao-li.com/Hao_Li/Hao_Li_-_publications_%5BExample-Based_Facial_Rigging%5D.html
2. This video introduces a method for generating facial blendshape rigs from a set of example poses of a CG character
3. I think this video conveys the idea, but can definitely be improved. I think it is a little too informal and the attempt to use humor has not worked very well. The video has many examples but the narrator does not point out what the key advantages of the introduced technique are.

sgallege April 7, 2011 at 8:00 pm

1. Motion Fields for Interactive Character Animation http://grail.cs.washington.edu/projects/motion-fields/
2. An alternative to graph based motion character animation, namely motion fields which is flexible simple to implement (I don’t know how true that is), easy to use and responds rapidly to user input .
3. I think over all the video is good, especially it make the point very well in the way they compare motion graphs and fields. The overlapping characters clearly illustrate how the motion fields are more responsive. The visuals, narrative and captions are well synchronized, and the quality of all the elements are good

Reid April 7, 2011 at 8:35 pm

1. Content-Preserving Warps for 3D Video Stabilization. http://pages.cs.wisc.edu/~fliu/project/3dstab.htm

2. This video tells the shows the viewer that handheld camera’s are shakey but to an extent this can be fixed. It compares commercial results to the paper’s results and gives a brief description of how it works.

3. The video did a good job of comparing the different methods as well as highlighting some of the problems with existing methods. The ‘technical’ description it gives is probably too vague to be of much use without the paper to back it up, but it gives a general overall idea of what’s going on.

4. The video conveys that this method is an improvement for amateur film makers over existing image stabalization. It offers the ability to simulate dolley shots where other methods only reduce the shake.

5. This video would probably work well to advertise the video, since it shows it’s results in comparison to commercial results. It has lots of video clips in it so people who wouldn’t understand the paper can look at it and get a sense of how it’s better than what else is out there.

6. I would say this video has good pacing. It is well synched with the narration, and the narration is long enough to give the viewer time to soak in each shot. It’s also large enough that the side by side shots aren’t too cramped together.

Reid April 7, 2011 at 8:48 pm

1. Rigid Body Fracture Sound http://www.cs.cornell.edu/projects/FractureSound/

2. This picture ‘shows’ the user they can make things sound like they are breaking.

3. This video tells the viewer what things breaking sounds like.

4. This video conveys their system does indeed sound like things are breaking.

5. This video would work to augment the paper. No narrations and very little text is displayed so it is mostly just the sounds of things breaking. As such it would need the paper (or at least the abstract) to inform the viewer what is going on. It does give many examples of their results and comparison’s though so if you read the paper the video would give you the aural sense of what it does.

6. The lack of narration and little or no text makes the sounds stand out a lot. Even the video clips shown are small which puts more emphasis on the sounds. The presence of the images of things breaking however is enough motivation to keep the viewer awake and engaged.

Reid April 7, 2011 at 9:31 pm

1. Piles of Objects (one of the readings from assignment 1) http://students.cs.tamu.edu/swhsu/proj/piles/

2. The video tells the viewer that their method can produce piles of objects comparable to a physics simulations.

3. The video does a good job of showing the system can make a pile of objects and just about nothing else.

4. This video conveyed piles of objects.

5. This video serves to augment the paper. Since it has very little text and no sound it is not an informative video. It has lots of video clips though, some of which highlight parts of the method they are using, but which wouldn’t make sense if you had not read the paper.

6. The video pacing seems a bit slow, they often show the same clip twice under the heading ‘our method again’. The lack of sound also makes it somewhat dull, a narration describing what is going on could have made it a better video. The showcasing of the results feels very sterile and to the point, nothing extra added.

Reid April 7, 2011 at 9:49 pm

1. Real-Time Deformation and Fracture in a Game Environment (again from assignment 1 readings) http://graphics.cs.berkeley.edu/papers/Parker-RTD-2009-08/

2. This video tells the user their method is capable of modeling realistic object damage in real time and with a wide variety of applied forces and materials.

3. This video does a very good job of showcasing the system, since it shows essentially end user interaction with the system in the form of captured game footage.

4. This video conveys the system is fast, robust, handles many materials, and is customizable.

5. This video does not describe how the system works, so it’s not stand-alone. It does however make excellent advertisement for the paper, perhaps the best advertisement you can get. Since it shows live game footage it has a lot of ‘wow’ to it which makes the viewer want to be able to reproduce what is shown. The fact that the game is also commercial and from an iconic franchise gives it lots of credibility.

6. I noticed that all of the live game footage is shown first, while the tech demo shots are towards the end of the film. This probably takes advantage of the fact that viewers are more likely to watch the whole thing if it grabs their attention at the beginning. It probably wouldn’t have the impact it does if the tech demo’s were first and the game footage at the end. Playing more to this point the first few shots are action shots as well, while the later shots are just the player fooling around with things.

Danielle April 7, 2011 at 10:25 pm

1. The Saliency of Anomalies in Animated Human Characters (http://graphics.cs.cmu.edu/projects/MMM/)

2. This video runs the viewer through a series of different experimental conditions used in a particular study and how those original motion capture segments were collected.

3. It’s hard to judge this video objectively, already having seen the accompanying talk. In the context of the presentation, this video provided a solid explanation of the different experimental conditions well-suited to a twenty-minute talk. However, out of context, I think it is much more difficult to tell what is going on. On one hand, it lets the viewer experience what the actual response is to the environmental conditions. On the other, it doesn’t give the user any context as to what the data is for and what they are supposed to be gaining from the video.

4. The video conveyed the actual different experimental conditions used in a study for purely supplemental purposes.

5. This work is intended to augment a paper. While the content of the video would be extremely difficult to effectively convey in text, the amount of information presented in the video is insufficient to explain the premise of the experiment.

6. One thing that was interesting about this video is that there was no narration — sound was reserved solely for the purpose of representing the study conditions. This had an interesting effect of having the user almost “take part” in the experimentation, but since I’m fairly certain that the study was not done zoomed in on the woman’s face, this addition of the zooming technique also emphasized the key changes that the study explored. Also, sampling the different audio tracks and splicing them together instead of showing the full scene each time helped introduce the conditions and context without boring the user.

sghosh April 7, 2011 at 10:31 pm

1) Rigid-Body Fracture Sound (http://www.youtube.com/user/DougLJames#p/u/0/nHH8N_lNZzI)
2) The video shows outputs of a few artificially synthesized, synchronized fracture sounds for rigid bodies under different input parameters.
3) All I understood from the paper is that yes rigid bodies produce different sounds when they crack under different circumstances – and this algorithm is damn intelligent in doing the same. But there was no explanation on the parameters and how exactly they affect the sound and neither is there a good comparison with real life examples.
4) The algorithm is able to synthesize the fracture sounds of various brittle objects when they fracture and actually synchronizes it with the state of that object.
5) Would it stand by itself? – no way. It’s kind of vague and incomplete.
Augment a paper – perhaps – if the authors have done a good job explaining the concept and the significance of the various parameters involved in deciding the sound (given the fact that most of the people involved in graphics related research areas would be very unfamiliar with sound related concepts )
serve as an advertisement for a paper? This is a difficult question – on thinking hard I think yes (kind of contradictory to the first answer). The very fact that this is a paper on sound synthesis in a graphics related conference will be a strong crowd puller.
6) Of all the videos I have seen so far (throughout the semester) this is the first one without any voice explanation – so it took me by surprise. Playing one sound after the other in quick succession and reading the accompanying text (much of which I did not understand) proved difficult for me to follow (I understood there was something different but I had no idea why so).
I am not sure about this (reading the paper would give a better idea) – I thought the paper is mainly about the sound aspect of the fractures – if that is the case I find no significance of the slow-motion replay of the dinner plate/wine glass cracking.

sghosh April 7, 2011 at 10:32 pm

1. Parametric Motion Graphs, http://www.youtube.com/watch?v=ZepNQHXAcCY
2. The video shows the results of how effective the algorithm is at interactive character controlling
3. I think the video does a good job explaining the usefulness of this method. There were lots of examples with different case. The author draws the attention of the viewer several times at certain important characteristics of the character. I liked the end clip with the hand held controller and the display in the background. That surely adds credibility at how easy it is for the user to control the character and how believable the output generated is. I would have liked it if there were some comparisons with other existing/previous methods.
4. It conveys that the parametric graph technique is effective at producing smooth interactive animated motion of characters that can be controlled by the user.
5. For all animation papers I guess it is a general case that a video showing the results is an effective way to show the salient point in the paper. The author does put in a lot of effort to describe as much as possible about the different motion types but not much about the inner details and neither about comparisons with other methods. So I guess this video is best when it accompanies the paper.
6. The background music was very annoying and distracting. For a good few seconds I was try to remember where I had heard this music and did not concentrate on the video. But apart from that it was well organized. The subtitles were useful as well.

sghosh April 7, 2011 at 10:32 pm

1. Performance based control interface for character animation (referred the first 2 videos here – http://www.cc.gatech.edu/~karenliu/Performance.html)
2. The videos show how easy it is for a virtual character to freely interact with the 3d world with just a few body gestures as inputs.
3. It was kind of odd that in the first video he speaks for the first 15-20 seconds and it’s pin drop silence for the rest of the minute. But apart from that both the videos combined are excellent at proving how easily and realistically they can control the virtual character. Sure they haven’t explained the method but the 2nd video in which we see the actual input and the computer generated output we get an idea how this algorithm works and how effective it is.

sghosh April 7, 2011 at 10:33 pm

1.Sneak Preview of New Motion Stabilizer for Adobe After Effects (from the new post) http://tv.adobe.com/watch/adobe-technology-sneaks-2011/sneakpreview-of-new-motion-stabilizer-for-after-effects/
2.The video shows how effective the new motion stabilizer in Adobe AE is in stabilizing jerky hand held videos.
3. I really liked the video. The author of the video gives a very clear and concise description of the utility of the algorithm. I am sure the intricate details of the algorithm will be complicated but even a layman can get some idea about how the algorithm generates ‘tracks’ and keeps track of jerks and negates the same. Overall I think it does an excellent task of telling the viewer that this is a good algorithm and probably compare fare better than other available commercially.

Leslie April 7, 2011 at 10:35 pm

1. Be clear about what the video is (name, type, link to where you found it on the web (preferably a page with the video, not just a link to the video itself).
Content-Preserving Warps for 3D Video Stabilization
http://pages.cs.wisc.edu/~fliu/project/3dstab.htm
2. In a sentence or two, explain what the video tells the viewer.
The video describes a unique video stabilization technique and compares it to existing algorithms. It illustrates the similarities and differences in both implementation and results.
3. Your overall impression of the video – how well did it make its point? How effective is it at conveying its ideas?
This video was excellent, and made its point very well.
4. What kinds of things about the work did it convey?
It touches on almost every aspect of the work. It gives some background, implementation details, and good and bad results.
5. Would it stand by itself, augment a paper, serve as an advertisement for a paper? Why?
I think this video stands by itself remarkably well. While most videos only show results, this video has the same type of structure as a paper—it starts with and introduction motivating the need for this type of video stabilization, then talks about background/ existing stabilization algorithms, then explains the methodology of the current algorithm, then gives results. It’s very comprehensive and easy to understand.
6. What did you notice about the video itself (in terms of its production)? Anything from quality of the audio, legibility, editing, pacing, … Just like with writing, a lot of what makes video good can be very subtle. Try to notice what kinds of things influence you.
I thought that, overall, it was nice-looking and professional. I thought that the narration was well synchronized with the images, and that probably had a lot to do with how so much information was presented in such a short period of time. I didn’t like when the narration stopped. In the project page version there is just silence, and in the Youtube version there is music that doesn’t seem to fit the video that well. Both are kind of strange.

Leslie April 7, 2011 at 10:35 pm

1. Be clear about what the video is (name, type, link to where you found it on the web (preferably a page with the video, not just a link to the video itself).
Parametric Motion Graphs
http://pages.cs.wisc.edu/~heckr/Research/pmg.html
http://www.youtube.com/watch?v=ZepNQHXAcCY
2. In a sentence or two, explain what the video tells the viewer.
This video shows the viewer some results of parametric motion graphs. It demonstrates locomotion, random motion, and interactive motion.
3. Your overall impression of the video – how well did it make its point? How effective is it at conveying its ideas?
It never really explains what parametric motion graphs are, but it does a good job of making them look awesome, and showing a wide variety of results and applications.

Leslie April 7, 2011 at 10:36 pm

1. Be clear about what the video is (name, type, link to where you found it on the web (preferably a page with the video, not just a link to the video itself).
Rigid Body Fracture Sounds
http://www.youtube.com/user/DougLJames#p/u/0/nHH8N_lNZzI
2. In a sentence or two, explain what the video tells the viewer.
The video doesn’t “tell” the viewer much. It shows a bunch of different things falling and breaking. Clips are divided up into sections that don’t make a lot of sense, at least to me.
3. Your overall impression of the video – how well did it make its point? How effective is it at conveying its ideas?
I was able to understand that it was about simulating the sounds things make when they break, but I didn’t understand any of the different parameters they listed along with the video clips. I would imagine that it conveys its ideas pretty well to someone who is familiar with the work.
4. What kinds of things about the work did it convey?
It showed tons of results. That’s pretty much it.
5. Would it stand by itself, augment a paper, serve as an advertisement for a paper? Why?
It definitely doesn’t stand by itself. I think the intention behind the video is to present as many results as possible, since there is no way to demonstrate or describe these types of results in a paper. However, it does a decent job of advertising the paper/research. After watching the video I was interested enough to Google the paper and skim the abstract to learn a little more about the project.
6. What did you notice about the video itself (in terms of its production)? Anything from quality of the audio, legibility, editing, pacing, … Just like with writing, a lot of what makes video good can be very subtle. Try to notice what kinds of things influence you.
It was strange that there was no narration, but I suppose that makes sense, since the viewer is meant to focus on the sounds that are a result of the research being presented. I don’t understand why they played silent, slow motion clips before they played the clips with sound (like the wine glass breaking). Maybe it was supposed to provide some reference, but I thought it messed up the pacing of the video.

Leslie April 7, 2011 at 10:36 pm

1. Be clear about what the video is (name, type, link to where you found it on the web (preferably a page with the video, not just a link to the video itself).
Spatial Relationship Preserving Character Motion Adaptation
http://www.youtube.com/watch?v=2Fua5M8gbto
2. In a sentence or two, explain what the video tells the viewer.
The video tells the viewer why it is important to preserve spatial relationships between a character and its environment, and then shows some results (both good and bad) of their approach (using the interaction mesh).
3. Your overall impression of the video – how well did it make its point? How effective is it at conveying its ideas?
It did a good job of showing that spatial relationships should be preserved, and the necessity of an interaction mesh. It conveyed its ideas very clearly.

Leslie April 7, 2011 at 10:38 pm

1. Be clear about what the video is (name, type, link to where you found it on the web (preferably a page with the video, not just a link to the video itself).
Parametric Motion Graphs
http://pages.cs.wisc.edu/~heckr/Research/pmg.html
http://www.youtube.com/watch?v=ZepNQHXAcCY
2. In a sentence or two, explain what the video tells the viewer.
This video shows the viewer some results of parametric motion graphs. It demonstrates locomotion, random motion, and interactive motion.
3. Your overall impression of the video – how well did it make its point? How effective is it at conveying its ideas?
It never really explains what parametric motion graphs are, but it does a good job of making them look awesome, and showing a wide variety of results and applications.

Danielle April 7, 2011 at 11:18 pm

1. Animated Transitions in Statistical Data Graphics (http://vis.berkeley.edu/papers/animated_transitions/)

2. This video tells the user about DynaVis, a perceptually-motivated animated system for exploring standard data visualizations. It also explained the results of a user study comparing the DynaVis techniques to standard animation and static visualization.

3. This video felt like a very clean and safe summary of the paper. It was very effective overall in conveying the results of the paper and giving a high-level introduction into the problem they were exploring and their solution.

4. This paper was very clearly divided into three portions of the work: the problem, the solution, and the user study/conclusions. This gave an overall summary of the work presented in the paper.

5. Because the video does present a clear beginning (problem intro), middle (solution), and end(experimentation/conclusions), it could stand alone. However, it does briefly mention certain aspects of the system not thoroughly explained in the video that make it a better accompaniment than stand-alone.

6. The video is very cleanly put together and generally focused on the presentation of the tool. It was really nice in the way the same visualization tool was used for the entire presentation. It put together a very coherent video. Also, visually showing the experimental conditions helped to explain the tests, like the anomalies paper, but was actually given in context (illustrating the narration).

Danielle April 7, 2011 at 11:27 pm

1. 2.5D Cartoon Models (http://www.alecrivers.com/2.5dcartoonmodels/)

2. This video introduces the viewer to the 2.5D cartoon modelling system. It also tells some of the limitations of the tool.

3. This video made its point well with respect to what the system does and what two of its primary limitations are. However, it did not do much in terms of explaining the details of the system, making it of more use to artists who may potentially use the system than the technical individuals trying to pull techniques from the paper.

Nathan Mitchell April 7, 2011 at 11:36 pm

1. Responsive Characters from Motion Fragments
Raw Mpeg4, AVI container
http://graphics.cs.cmu.edu/projects/rcfmf/
http://graphics.cs.cmu.edu/projects/rcfmf/rcfmf.avi

2. The video demos a interactive animation controller that is influenced by prior user input. It compares the results of the controller to an offline optimal solution, a steady state controller, and a greedy controller.

3. In general, I was not very impressed. Partly this may be a personal bias as this was the paper we worked on for our project, but the overall differences in the animations produced by the various controllers ( apart from the greedy approach, which was visibly poor ) where similar enough that without the video’s highlighting the differences, I might never have noticed them. While I have a good understanding of what is going on behind the scenes, I don’t think this video gave a flattering impression of that work. Even the animation choices that the researcher’s picked leave much to be desired. They advertise this controller as useful for games, but most modern games have much more natural looking motion. The figure in the video looked stiff.

4. Again, my knowledge of the paper’s details allows me to answer this question. The video is very light on the actual work being done. While they mention the player’s control history briefly as an influence, they do not discuss the use of many small motion fragments and that the policy is generated by a reinforcement learning technique. The closest they come to that statement is where they mention the importance of considering all transitions, but they never explain that it is because of the reinforcement learning algorithm.

5. I personally see this video as augmenting a paper, perhaps as an advertisement. I think it is nice to be able to see the results of what they discussed in the paper, but I also think that the title of the paper is more of an advertisement than this video was.

6. The video was of good quality with enough resolution to see clearly. Also the voice work was clear and understandable, even if slightly uninformative. I think what caught my attention the most was the unusual choice of backgrounds. The wild rainbow patterns in the air were very distracting and, until I forced myself to focus on the character, disrupted my ability to make out the details of the animation. I am concerned that the choice of visuals may have been purposeful to pull attention away from any flaws in the character motion.

Danielle April 7, 2011 at 11:36 pm

1. Responsive Characters from Motion Fragments (http://graphics.cs.cmu.edu/projects/rcfmf/)

2. This video introduces the motion fragments technique and compares it to other fragment selection models.

3. This video is not as convincing about the actual strengths or weaknesses of the techniques mostly because the audio is so poorly synced with the video. However, it does make the comparison points reasonably well through juxtaposed and labelled clips.

adrm April 7, 2011 at 11:54 pm

1. Retargeting Motions to New Characters (http://www.youtube.com/watch?v=Vn-vVzMGgec)

2. This video tells the viewer what types of problems arise when applying motion capture data to skeletons of a different size than the original motion. The video then shows some example retargeted motions using the methods from the paper, but does not explain how the method works.

3. The video does a good job of showing compelling problem cases, and convincing the viewer why intervention is needed. The presented solutions work well, and the video gives the viewer an idea of the capabilities of the approach without going into any technical details.

4. It conveyed what problem the paper is trying to solve. It also showed a whole bunch of examples, and even a few failure cases. It however, did not really give much insight into how the results are achieved. (The what without the how)

5. In my opinion this is an advertisement for the paper, mostly because of reasons explained in 4. The lack of a technical explanation, even at a high level, but the presence of compelling examples makes me want to go read the paper.

6. This video was well produced, examples where clearly staged and chosen. I am personally not a fan of the music, and the voice is a bit saturated, so the pauses between clauses are very noticeable.

Nathan Mitchell April 8, 2011 at 12:01 am

1. Magic Board (1)
Raw Mpeg, AVI container
http://pages.cs.wisc.edu/~michaelw/index.cgi
http://www.cs.wisc.edu/graphics/Video/Gleicher/MagicBoard_MPEG4.avi

2. This video demos a system to replace information seen written on traditional chalkboard in a video recording with typed text, high quality images, and videos.

3. The results were impressive, though it took me a moment to realize that what was shown was a post-processing of the original video where the presenter wrote on the board. Still, the added artifacts impressively handled being correctly obscured as the presenter moved in front of them, giving the appearance of a multimedia system, even though it was a chalkboard.

4. The video was fairly light on implementation details, but it demoed the important ability to insert cleaned up text and multimedia seamlessly into the video. There was a brief explanation of the workflow, which made it clear that the act of choosing the multimedia and replacement text was still being done by a human, while the content identification, regions, was being done automatically.

5. I see this paper as a great advertisement. There are not enough details to understand how it works, but the results make me want to read the papers to find out exactly how it all works.

6. Given that this video was a live action performance, there is the additionally component of a real actor ( in this case the researcher ). The camera work was good, the scene well lit – but I felt the acting could have been better. The voice was a bit too quiet and the presenter stumbled occasionally during the speech. Overall the content allowed me to ignore most of these things, but there is definitely room for improvement.

adrm April 8, 2011 at 12:04 am

1. Spatial Relationship Preserving Character Motion Adaptation (http://www.youtube.com/watch?v=2Fua5M8gbto)

2. This video shows examples of problems that arise when characters are interacting with an environment or with other characters. It briefly explains their solution, and shows examples of how the method performs.

3. The video gets its point across very well, presenting the problem and solution concisely, while at the same time being able show a variety of examples and cases where the method does not perform well.

4. The problem, how they solved it, where it works, and where it does not. At varying levels of detail, but this video explores all of these aspects of their work.

5. I would say this paper stands on its own for most audiences. Unless one is looking to implement their method, the video does a very good job of explaining their approach at an appropriate level of technical detail.

6. The examples are well chosen and staged, but there are a few long pauses in the audio that a bit noticeable. Also the audio quality is a bit low.

adrm April 8, 2011 at 12:14 am

1. Interactive Generation of Human Animation with Deformable Motion Models (http://www.youtube.com/watch?v=v97xMxl-4eU)

2. This video presents a system that allows users to “sketch” different constraints on motions. The system shows plenty of examples of input and results, some of which are very compelling but there is not much technical detail on how the system work.

3. It is a very good video that shows off what the techniques in the paper can and cannot do (but not how), but it does not stand on its own (more of an advertisement for the paper)

Nathan Mitchell April 8, 2011 at 12:20 am

1. Parametric Motion Graphs
Windows Media Video, WMV container
http://pages.cs.wisc.edu/~heckr/Research/pmg.html
http://pages.cs.wisc.edu/~heckr/Videos/PMGDemo.wmv

2. This video explains a system to control a character’s animation in real time using parametric motion graphs. The method connects motion segments together almost seamlessly, allowing for random walks in addition to motions with parametric constraints.

3. I thought the video was well done. It demonstrates clear examples of the system’s capabilities without any visible artifacts that I could see. Additionally, the voice acting was clear and provided a good explanation for what was being shown.

adrm April 8, 2011 at 12:21 am

1. Example-Based Facial Rigging (http://www.hao-li.com/Hao_Li/Hao_Li_-_publications_%5BExample-Based_Facial_Rigging%5D.html)

2. This video presents a system that can transfer a blend shape based rigging model of a generic face to any other facial target. Adding more examples increases the quality of the rigging transfer.

3. Well does video that aims to tell the whole story. Its a bit more on the technical side (need to know a bit about rigging and blend shapes) and is a bit fast paced, but ti shows off the system very well.

Nathan Mitchell April 8, 2011 at 12:31 am

1. Rigid Body Fracture Sound
YouTube Video
http://www.cs.cornell.edu/projects/FractureSound/
http://www.youtube.com/watch?v=nHH8N_lNZzI&feature=player_embedded

2. This video provides examples of the researcher’s sound rendering system for rigid bodies (both for impact and shattering). It also provides comparisons of the rendered sounds to the actual sounds of real life objects under the same conditions.

3. This video was very impressive to me. While I lack the technical expertise to determine if the sounds being rendered are completely correct, I certainly would believe them if I could not see the rendered images. In that sense, the video has me convinced of the effectiveness of their technique. However, I do feel it is entirely supplementary to a paper as there is no voice over describing what is going on and there are no details as to how the sound generation is done.

raja April 8, 2011 at 2:15 am

1) Rigid Body Fracture Sound
Webpage: http://www.cs.cornell.edu/projects/FractureSound/
Video Link: http://www.youtube.com/user/DougLJames#p/u/0/nHH8N_lNZzI

The video shows the sound produced when brittle objects like ceramic plate, wine glass and glass windows are dropped/broken.
The author(s) don’t speak at all, apart from going “how did they do this?”, the viewer learns nothing.
There is some text, which I think has to do with filtering frequencies below a certain level, and figuring out the kind of sound produced.

Only that they generate rigid body fracture sounds in some magical way.

The video makes you want to know how they did it, but doesn’t tell you anything about it.
So, as an advertisement, yes, it worked. I did check their web page.

Definitely needs a paper to tell you how they did it. Its a good video from a “go to my web page” sense, but they could have at least mentioned a few technical terms that served as the basis of their idea.

Audio quality excellent (understandable as it really IS about the audio). The examples were interesting, but dropping plates over and over again w/o really telling anything about what is changing (ok, the text didn’t really help me..) is not useful from a viewer perspective.
Overall, 3/5 only because of the coolness quotient.

raja April 8, 2011 at 2:15 am

2) Spatial Relationship Preserving Character Motion Adaptation
Webpage: http://homepages.inf.ed.ac.uk/s0576323/projects/mesh/
Video: http://www.youtube.com/watch?v=2Fua5M8gbto

The video presents a method that allows animators to easily edit and retarget movements involving close interactions while preserving the spatial relationships of the scene components (this the first thing the authors say..)

Excellent. The video describes what they’ve done, where exactly it is useful, a little about how they’ve done it, a key idea that previous methods in retargeting didn’t use (previous methods preserved joint angles, while they preserve the spatial relationships using an interaction mesh) and also talks about some limitations of their method.

Using the interaction mesh to capture close interactions which is useful for retargetting motion to different characters.

The video really augments the paper and works as a fantastic advertisement. I believe in general that most animation/graphics papers can’t really describe much technicality in the video. I feel that hinting some of the ideas used boosts the advertising potential rather than only showing the results (however cool it may be..).

The production was very good. The speaker was clear, the pace of the video was easy to follow, good use of colors to show/differentiate characters/objects and a variety of applications were shown.
Overall, I’d give it 5/5.

raja April 8, 2011 at 2:15 am

3) Group Motion Editing
Webpage: http://mrl.snu.ac.kr/research/ProjectGroupMotionEditing/index.htm
Video :http://mrl.snu.ac.kr/research/ProjectGroupMotionEditing/GroupMoEdit_divx_mp3.avi
http://www.youtube.com/watch?v=xxCFvzPkjUs

The video presents a method to perform group motion editing which preserves the neighborhood formation by manipulating characters of the group. It also takes into account collision avoidance, group motion merging and obstacle avoidance to come up with the resulting group motion.

I think the previous answer (i.e. above) captures a lot of what I’d want to say here.

Brilliant. The video describes what they’ve done, where exactly it is useful, the extra features that it takes into account, the real-time editing capability and some cool applications.

The video serves as an excellent advertisement, while giving a good preview of what the paper is about. It mentions some of the ideas/methods used and keeps the user engrossed throughout.

Excellent pacing, clear audio, interesting idea and superb examples that will definitely make you check out their paper.
Overall, 5/5! No questions asked!

raja April 8, 2011 at 2:16 am

4) Synchronized multicharacter motion editing
Webpage: http://mrl.snu.ac.kr/research/ProjectMultiMotionEditing/
Video: http://www.youtube.com/watch?v=wiRmVsTarAI

The video describes a multi-character motion editing method that allows the user to manipulate synchronized multiple character motions interactively.

Superb. Makes its point really well, effectively conveys the problem they’ve tried to solve with neat examples and a great user interface that makes you want to read about it.

It mentions the Laplacian curve editing method for both editing both the path and timewarping. It describes extending the motion, either by making it longer in length and time and adding the idling motion sequence to achieve that.

The video again is a great advertisement and preview of what their paper is about. Completely sold me on it!

Great examples, clear audio. Very nice user interface, interesting problem, great application examples
If I absolutely have to find fault, it would be that the audio is a little sluggy (the speaker pauses a little more than necessary between words).
Overall, 5/5 again.

{ 2 trackbacks }

Previous post:

Next post: