I always like to start out with a few seminal papers. Stuff that’s so old that its hard to imagine, but still worth having seen. Looking at these papers are mainly a history lesson.
This list is the same as it was long ago, and it was re-iterated by the appearance of a “seminal papers list” from SIGGRAPH. (You might want to look at it, since many of the other ones are historically interesting too. I can’t say I’ve read all of them).
Several of the papers on the animation list, we will get to later in the semester (faces, animation principles, flocks). And the legged animation paper we will replace by another early paper by the same authors.
Read these two papers – they are really old, and the details don’t quite matter. You should be able to get the general gist quickly.
A System for Computer Generated Movies. E. Catmull. ACM ’72 Proceedings of the ACM annual conference. (ACM DL page).
Interactive Skeleton Techniques for Enhancing Motion Dynamics in Key Frame Animation. N. Burtnyk and Marceli Wein. CACM, October 1976. (ACM DL Page) (The most famous film from this effort, The Hunger, is online). We might watch some of it in class – otherwise, have a look at it.
And, in a comment to this posting, add a note explaining (these are related):
- Why do you think I asked you to read these ancient papers?
- What can we learn from them for the “modern” era?
Please do this before the class discussion on Friday, January 28th.
{ 14 comments }
You asked us to read these papers in order to facilitate an intuition as to why/how modern animation techniques exist. By understanding the thought process of ages past, we can clearly analyze their flaws given the reference of today’s methods. Ultimately, we should apply this same objective critique to modern methods in order to identify their limitations with the intent of building/improving upon them (as we have using these seminal papers).
Considering that the underlying concepts outlined in these papers provide the foundation for current animation, I think we can derive two main lessons:
1) The ideas presented from the work of others can serve as a point of entry or reinforcement to your own efforts.
2) Don’t accept existing standards as infallible. It seems that the ‘standard’ pipeline has become so engrained in CG that even if a massively revolutionary procedure was presented today, it would take many years to actually adopt it. Although a certain level of standardization is needed, over reliance on this standard can stifle progress. These papers prove the value of experimentation in CG and the idea that non-traditional methods should be considered viable alternatives; there is an off chance that they could change everything.
Even though the techniques in these papers are no longer state-of-the-art, the papers themselves are still valuable. This is in part due to their historical significance (as mentioned in the assignment), but they also give some insight into how computer animation has developed as a field.
I think the most interesting parts of these papers are their inspirations for the techniques they used, and their predictions for the future. For example, Burtnyk and Wein write that the idea behind their technique for controlling motion between key frames is very similar to a traditional animator drawing stick figures to give a rough idea of what a character’s motion should look like. I thought it was interesting that the authors were facing a problem similar to what animators had faced (trade-off between control of motion and conservation of resources), and so developed a solution analogous to the one animators had already discovered (although the technical details were of course very different). And Catmull, in the “Filming” section of his paper, writes that “A great deal of research still needs to be done on the creation and manipulation of three dimensional objects” and “[The rate of change of motion] is a non-trivial problem if a nice-looking action is desired.” These are both major topics that have been heavily studied since 1972, and have seen tremendous progress.
These papers are a reminder that the computer animation research being done now is a product of what came before it, and will influence what comes after it.
I believe the primary purpose of reading these papers was to make us appreciate the animation tools we have now. I may grumble about how hard it is in Maya to figure out how to navigate the labyrinthine menu structure to create some NURBS surfaces and give them skin texture to make a hand, but at least I don’t have to shave the back of my hand, create a plaster mold of it, and draw polygons directly on the mold as Catmull did.
I found it very interesting, especially with Catmull’s paper, to see what has and has not been solved in the past 40 years. Figuring out how to use a sequential programming language to create simultaneous overlapping action isn’t even close to the conceptual hurdle it apparently once was, but animations of physical bodies such as cloth, which Catmull mentions briefly as being “difficult to model”, are still open research questions. Also, his organization of objects into bodies using a tree structure seems to be the forerunner to the use of articulated figures in animation today.
Most importantly, what I take away from these readings is a line from the Catmull paper: “For making movies, it is more desirable to be a director than a programmer, recognizing, of course, that being a good director is still a lot of work”. In 1972, the biggest obstacle to making computer animations was the complete technical lack of tools. Today we are getting to the point of having all the tools, but the concern is figuring out what to do with them. Catmull recognized that even once we have all the tools, making really good animations – being a director – will still be difficult (and interesting).
1) Why read these papers?
Going into this class many of us have experience with various animation or graphical systems, even if it’s just the OpenGL programming in 559. Some of it we know “from scratch” (we know enough of the math behind it to make a stab at implementing from the top down), but a lot of it we don’t. By the end of the second week of an animation class we are in roughly the same spot as the paper authors: we don’t have a clear idea of how the low-level things we take for granted (interpolation, rotation, skeletons, &c.) work. We’re in a slightly better position than them because we know for a fact that these things do work, and they do work well. But a lot of the specifics of how and why they work are hidden behind layers of GUI and optimization and interactive tools. Just like the nlerp and slerp comparison in the last readings, insufficient understanding of the underlying structure and function of different graphical systems leads to people making non-optimal choices, using magical thinking (I put this thing in my code and it worked, so I’ll just stick it in whether it’s appropriate or not).
2) What can we learn from these papers?
i) How can we do cool things computationally at a very low level of resources and infrastructure?
ii) What metaphors from 2D animation make the transition to 3D? If they don’t make the cut, what metaphors can we adopt instead?
iii) How did the field evolve, and how can we predict where it’s going? To what extent are we still in the shadow of these early papers?
iv) What mistakes can I avoid, given the shortcomings and false optimism of the papers (pulled hair excluded)?
In animation, and graphics in general, there are almost always multiple ways to do the same thing and there are almost always problems with each method. In general, solutions to these problems can lead to more problems and eventually you get to the point where any paper you’re reading on Animation or Computer Graphics will really be an attempt to solve a problem that has come about due to the way that previous problems where solved. Reading “ancient” papers is a good way to discover the foundations that modern papers are built on. Often, it is only possible to really understand the modern solutions when the base of the problem is understood.
What we need to understand from these papers is what has proven the test of time. For example, the MOP language is probably not used by any humans in modern times. But it might form the basis for file formats in standard modeling and animation packages. The methods in the Interactive Skeleton paper look a lot like Texture Mapping Coordinates.
It’s also useful to see which problems where tackled first, what solutions where proposed for them and in what domain they where considered. For example, the Skeleton paper was considering problem solutions in the 2D realm, which at that time would have been where much of that actual Animation work was being done.
Finally, reviewing old papers gives us a chance to look at solutions to problems that my rear their head again. While 3D animation has been the norm for the last 5 years, 2D animation seems to be making a comeback and it might be the case that the solutions to 3D problems might aid in problems that arise in 2D animation.
1) Why read these papers?
As a student of Computer Animation I believe the main purpose of reading these papers is to understand the evolution of this field – how much it has progressed from its humble beginnings some 4 decades ago and to be able to better appreciate the small but significant contributions of many along the way. Today any novice can watch a ‘tutorial’ video on youtube and figure out how key-framing can help him to make a smooth animation. But he has to realize that back then when Gouraud had just developed his shading method and when the concept of hidden surfaces was just realized, this was no easy task.
2) What can we learn from them for the “modern” era?
In the modern era, as we strive to make things even more believable and available on a smaller form factor device- we will be faced with newer problem. Sometimes looking back can help one to better understand the root of the problem or take a better approach. Carefully evaluating the work of many researchers over the decades on similar problems will help a lot there. Often one might even realize that an earlier simplistic method is actually much better than the present much ‘cooler’ approach. Apart from that, these papers also give a sense of direction to the computer animation industry as a whole – 4 decades back we were striving to get 2D animation look ‘perfect’ but then the desire to make things realistic has enabled us to move to a higher dimension. So the problems we face now now will in a way decide what movies and games we play in a few years.
When reading these papers, it is important to keep in mind the three categories that the problems of the author’s fall into. First are the problems that are no longer relevant: the somewhat humorous warning about shaving and plaster sounds very odd to modern readers. Likewise, the physical setup they used in the first paper seems very archaic as we can now record movies directly from the computer.
Other problems, including the rendering rate, have been solved through algorithms but also pure brute force. The computational power available now to computer animators puts to shame anything the authors had available. Though, interestingly, the hand animation from the first paper, while simple to the modern critic, actually looks rather good. This demonstrates how much of animation depends on motions and positions, and not perfect visual realism.
However, there are problems that are not solved, or not solved satisfactory. Looking at the description of language requirements, we can see that these needs are not that different from modern needs. And some of the points are still areas of active research and development. Our ability to do animated physics and fluids is still mediocre at best. Additionally, what tools we do have are difficult for anyone but trained professionals to use well – indicating how much of the process requires a human operator.
On the other hand, the authors in both papers have done a oddly prophetic job of hitting many major areas that are considered core in graphics and animation today. The concept of skeletal animation is now a primary method of animation. And as the first paper points out, good animation of the hand would not have been possible without realistic shading – something that is continually being researched today.
In total, the major point of these papers is not to learn some rare item of arcane knowledge. It is to understand that the field of computer graphics and animation is not so much revolutionary but evolutionary. Between these two papers, many aspects of modern graphics can be glimpsed in their infant form – still early and unformed, but easily recognizable.
1. I think these papers were assigned so that we would understand where animation was, and how it got to where it is today. Reading these papers certainly gives a sense of how far modern animation has come. It shows these papers are in a sense at the core of modern animation. Key framing and skeleton based character modeling are an essential part of modern software such as 3DS Max.
2. It also shows that a lot of modern problems are extensions of older problems. In the paper on cg movies, the primary difficulty they had was modeling the hand. While modeling a hand in today’s world is a trivial task, we’ve simply upped the ante and are trying to model more complex things, and more quickly, giving rise to many of the computer vision techniques that are being used to acquire model data from video. To sum it up, we can learn from these papers that helpful contributions to animation will likely entail improvements that A) make it easier to make models and/or B) make it easier to animate models
1. Reading these papers is a valuable experience because they lend an appreciation of the evolution of animation techniques. These two papers in particular give an interesting view of the animation world before and after key framing, and show how a concept that we take for granted, and which we think is so simple now, was once revolutionary. Before key framing, if the Catmull paper is representative of the time, mathematical formulae and large quantities of data were used to drive animation sequences; after key framing, this is no longer necessary by virtue of being able to use interpolation.
2. Both of these papers are concerned with the creation of tools or new methods of animation, and both are concerned with bettering our ability to produce animation that is realistic. However, I think that now we also concern ourselves with techniques and tools that produce artistic effects. I approve of this branching of efforts because I don’t see a pressing need to improve animation to the point where it is indistinguishable from, say, high resolution video (I believe there is a term for this singularity but cannot remember it). In any event, the invention or refinement of new techniques is at the core of computer animation.
These older papers help show some of the ways current techniques were discovered and built on top of older ideas. They could also give ideas about how to build further on current techniques, as previous examples of successful improvements are likely good models for the future.
Some of the problems presented in these papers are things we might now take for granted and then ignore the prior work that made it possible. Animating a hand through a computer may not be a difficult task any longer, and some of what still seems difficult today likely won’t be so difficult at all in another thirty years. So it’s important to learn not just about today’s problems in animation but also ways to solve problems that will come up in ten or twenty years from now. However, there are also things in the papers such as key frames for animation which are still useful for animation today. So some concepts seem like building blocks to larger ideas, while others stick around and are independently useful.
Why did we read these ancient papers?
As present day students, many of us in the class have been exposed to computer animation for most of our lives, either through tv, movies of video games. This causes us to to have very different perspectives and expectations about what computer animation is. By reading these papers, it forces us to think about the problems that researchers faced back then. This, coupled with our “current” day view of animation, can give us a glimpse into the evolution of the state of the art, which might help to make sense of some “arbitrary” design decisions that we take for granted.
What can we learn from them for the “modern” era?
“Those who don’t learn from history are doomed to repeat it.” We don’t want to keep reinventing the wheel. If we understand the evolution of the problems in animation, then perhaps we can also try to understand their solutions. Sometimes it’s too easy to concentrate too much on the details and loose the over all picture.
First of all, I realized that Catmull’s paper has acknowledged Ivan Sutherland, whose “Sketchpad” paper in 1963 was in true sense seminal and visionary (and one of the most elegant and original PhD thesis to read).
The other two papers by i.e. Catmall (1972) and Burtynk (1976) provide an insight into researcher’s grand vision and creative ideas when computer systems were evolving. It is truely unbelievable that they had workable solutions when the computer RAM was less than 64KB and disk space less than few megabytes.
Both the papers were written before “Quaternion” became standard in animation community and the paper gave some insight into problems associated with rotations in early systems. Catmull’s paper touched upon “SceneGraph” and “Concurrency” issues which are even more relevant today.
Butynyak’s paper provide a deep technical insight into the development of “Key framing” and after reading the paper, I feel that the time is frozen and very little technical ideas have evolved during 1977 -2010. In fact the picture on page 568 is so much similar to some of latest SIGGRAPH paper.
The biggest think to learn from these papers ( as a researcher ) is to explore new ideas and implement them even when we think the current technology and available resource s are limited.
1. Reading these ancient papers gave me a sense of what was possible in the 70’s itself in the field of graphics and animation. Its important to remember that mathematics was so well developed even then, and it was really about mathematicians using a subset of ideas in the then new field of computer science.
Catmull is a legend in graphics (student of Sutherland) and utilized the smooth shading and visibility solutions and some neat data structures to come up a hand animation that looked pretty good (esps for that time). The mold on the hand to actually create the polygonal representation was hilarious!
The skeletal animation concept is another beautiful one forming the basis for animations today. Again, the importance of interpolation stands out. The stroke interpolation idea (used in 2.5D) actually has its origins here! Some of the general statements in these papers are quite prophetic and amazingly still problems today (cloth!).
2. I wouldn’t want to say “what happened to research, if these were possible in the 70’s.. A lot of research since has allowed the widespread use of graphics. Its useful to remember that Quaternions was a 1800’s idea and only in the 90’s (?) did scientists think of using it widely in graphics in rotations. The problems we face now are many-a-time “make run faster”, “make look better” kinds.
These papers are important for understanding the fundamental techniques used in animation. Since they are the ‘seminal’ papers in presenting these particular ideas, they focus on explaining these ideas in full detail, including analyzing any drawbacks of the methods. Furthermore, they present methods that atttempt to simplify the automation of movie making through treatment of physical objects as data structures (i.e. skeleton control and object hierarchies).
Also, these papers help define the key ‘steps’ in animation. For instance, the Catmull paper laid out the types of controls necessary for successfully developing a film making language, including collision detection and motion paths. These sorts of insights set the groundwork for the development of animation as a field.
To tie into the current lecture sequence, both papers also highlight the necessity of rotation and interpolation to animation development. The use of interpolation in both papers gives a practical grounding to the work we’ve studied thus far in class and also presents the ease of interpolation in computer-based animation as the key win in using automated methods over traditional hand animations.
Through these papers, we can discover the advantages and drawbacks of traditional ways of thinking about animation implementation. While current papers may take the shortcomings of these techniques for granted, looking at a focused analysis of these methods provides an analytical understanding of what’s going on in these foundational methods and can apply this understanding to more modern methods.