|home | projects | opinions | contact  |
|  ray luo : projects : penimation|
CS 184 Penimation Final Project
| Links | Current Status | Project Proposal | First Deliverable | Second Deliverable | Third Deliverable | Project Wrapup |
Download Penimation Updated Version(Fixed some bugs that caused out of memory error, etc)
Check out the Third Deliverable for some screen shots
The updated version of Penimation is here as a zip file (2.5MB). I added some widgets and got rid of a bunch of animation frames so that out-of-memory exceptions don't occur any more. Also fix some other bugs and added simple stroke preference support. Finally, I changed the bouncing ball and space flight animations. Note that the space flight still takes a while to load (a progress bar should be implemented in the future).
The latest version of Penimation (as of December 10) can be downloaded here as a zip file (2.9MB). No installation necessary; easy to use; easy to build. Read the README that comes with the zip file for simple setup instructions. Penimation requires Java 2 JDK 1.3 to run. I haven't tested it for UNIX systems, but it works fine on any PC with JDK 1.3.
Penimation comes with C++ code with Renderman binding for generating animation sequences that come with Penimation. You'll need Visual C++ and BMRT to run and build this portion of the application.
Penimation comes with an unmodified redistribution of the binaries for the Satin Version 2.2 Toolkit for your convenience. See the copy right statement in the classes directory. You can also get the full version of SATIN from GUIR. Only two isolated areas of code uses SATIN. You can modify it easily to use your own gesture recognizer. In particular, I did not use any of the SATIN core or optional library classes beyond Sheet and RubineInterpreter. You can replace Sheet with your own glass pane and RubineInterpreter with your own gesture interpreter if you wish. On the other hand, the good folks at GUIR of Berkeley spent a lot of time making the toolkit powerful and easy to use, so you may consider taking advantage of it.
It is often difficult to manipulate moving pictures is a precise and intuitive way. One problem is that video button controls don't tell us enough about how fast the animation is moving, where in the animation are we interested in, and how we can manipulate animation sequences. Hence the development of Penimation.
This project is a case study in developing one or more animation sequences and a gesture-based graphical interface for viewing and manipulating these sequences. The animation will be developed using BMRT 2.5 to ensure a high degree of realism. I will write an interactive tool for generating these sequences, although they will not be part of the final user interface. The animation viewer will be developed using Java Swing, with gesture recognition support coming from the SATIN toolkit. The gesture set will be trained using GDT, the gesture design tool.
Most of the work in the Penimation project focuses on methods for generating animation sequences that are optimized for gesture-based interaction. The animation will be supported by an interaction tool that supports gestures for playing, fast forwarding, and rewinding the animation at different speeds as determined from the gestures. The user will also be able to zoom into a region of interest, zoom out, manipulate a set of sequences (moving, deleting, duplicating), and possibly making changes to the way the animation is displayed or the animation sequence itself, all supported by informal pen-based interaction.
Penimation will experiment with novel interaction techniques in viewing and manipulating animation sequences. Gesture-based interaction techniques will be used to support precise yet intuitive animation control.
By November 27, I hope to finish most of essential components of the GUI, using an example animation to demonstrate it. This includes getting gesture recognition to work, displaying the animation using Java at an acceptable speed, timing the animation sequence, building the GUI, and modeling user interaction. By December 4, I hope to finish creating one extensive animation sequence, or two or more simpler animation sequences. The RenderMan code (given in C-binding form, using Visual C++) will have to be completed. This includes rendering the scene, modeling the objects in space and time, achieving a compression format, negotiating with the GUI, and improving realism using shaders, texture-map, motion-blur, and other utilities provided by BMRT.
The Penimation project focuses on animation, novel pen-based user interfaces, modelling in animation, and to some extent, rendering of sophisticated scenes using lighting and shading models. Approximately 30% of the work will be in animation, 40% in user interface (aka other), 20% in modelling, and 10% in rendering. In particular, the work in modelling and animation accounts for a total of 50% in creating the animation sequences, with another 10% for shading, texture, etc. There may well be more work in the 40% in user interface due to the use of gesture recognition, user interaction, and timing sequences. This may take the user interface component to 50% while subtracting from the modelling component.
Update on Deliverable, November 27
The essentials of the pen-based GUI has now been completed. I used an example animation from BMRT to test the rendering, and performance appears to be acceptable. Basically, I used a Timer object in Java to fire action events to my animation engine, which renders the the animation sequence from file. There are still problems with the way the animation proceeds. For example, I get a noticeably discrete animation sequence for certain frame rates.
The gesture recognition is acceptable. Some commands are easier to to detect than others. File command, for example, tends to be missed more often than others. The essential commands, however, perform rather well. The stop command, for example, is always recognized and produces no noticeable latency between gesture interpretation and command execution. The following commands have been implemented, both as gestures and as button/icons: file menu, play, reverse, rewind, fast forward, stop. There will be more to come. While the skeletal code can be demonstrated right away, the details (such as creating jpeg button icons, maximizing performance, acheiving file compression, and embedding sound) have not been dealt with. Instead, I began the animation component by writing some template Visual C++ code to be used for generating the animation sequences.
I came across some other ideas that I thought were interesting if I could incorporate them into Penimation. First, there ought to be an intuitive way to edit an animation frame or add components to a collection of animation frames. It would be useful to come up a novel way to interact with the frames within the GUI so that animations can be changed dynamically using a pen. Second, the animation modelling ideas presented in CS 184 were very interesting. I will incorporate some of the ideas (such as modelling a water fall or explosion) in my animation. I project to spend most of the rest of time rendering and modelling the animation sequences. The interaction with the sequences will also be high in the list of agendas.
Update on Deliverable, December 4
An early version of Penimation can be downloaded. I have developed two simple animation sequences, one dealing with a bouncing ball, and one emulating a space flight. Lighting, texture, and other rendering considerations have not been applied to the frames yet. Simple models for animating the fully developed sequences are complete. For the final deliverable, I need to make the model more sophisticated (dents for bouncing ball, background for space flight, etc), apply custom lighting, textures, and shading, and make the images more attractive. The basic framework is complete.
Meanwhile, I've added a file chooser to the GUI and eradicated some non-standard behavior. I've tried to change the timing mechanism to yield a smoother animation display. Loading animations, however, still took a lot of time. I need to make a progress bar to keep the user busy while I load in the animation frames. I also need to find a way for users to draw on the animations, so that the gesture set will be less limiting and more powerful.
The first priority at this point is to refine the animation model for the bouncing ball to include damping and slow down frames for which the ball is near a local maximum. Refine also the space flight sequence to enable movement in three dimensions with turns and rotations. Then complete the renderings with textures and custom lighting and shading. Finallly, refine the interface to allow for user pen interaction directly with the frames and speed up loading and execution.
Final Update on Deliverable, December 10
When the program starts up, three frames are shown: the main Penimation animation window, the toolbar, and the palette. The latter two are for gesture commands and editing options, respectively. The main window is where all commands and editing ultimately being drawn. The two modes supported by the system are gesture commands (left mouse button) and editing scribbles (right mouse button).
Here's a screen shot of the toolbar. Each icon corresponds to a particular animation snippet for the gesture corresponding to the action specified by the action. The black dots in the pictures shows where the gesture should begin. The gray paths indicate the directions appropriate for each gesture from the starting point. Clicking on the buttons also invoke the action, but button actions are more limited. For example, we can't indicate how fast to play an animation with simply a button press. We can, however, determine the desired speed from the "play" gesture, by taking longer strokes to mean a faster animation speed.
When the pen (or mouse) rolls over an icon, the animation begins to play, showing the user how a gesture is performed. Note that the mouse is hidden in this screenshot. In this case, the mouse is over the rewind icon, where the animation has already shown half of the frames. The animation starts off slow and accelerate to match the view of a natural human stroke.
Here is a shot of a rewind command made on the Penimation window, with the space flight animation playing in the background (Mouse is hidden). The thick lines and gray color indicates that the stroke is a gesture that will be interpreted by the system. Gesture commands are made using left mouse drags (or unmodified pen strokes). Some gesture commands can also be performed using the tool bar frame.
Here is a view of an edit on the color spheres animation (comes with BMRT, by Larry Gritz). You cannot see that the mouse in picture has changed to a pencil/paper, clearly indicating which mode the user is in. Editing is done using right mouse drags (or modified pen strokes). Note also that the color and stroke of the edits can be changed using the palette frame.
Here is a look of the palette in the latest version. I added support for stroke size.
The next screen shot shows a stop gesture on the space flight animation sequence, one of the animations that comes with Penimation. Note that this particular sequence takes a long time to load. A progress bar should be implemented.
Here is another shot of the entire interface. It also shows me drawing on the bouncing ball animation sequence. With time, I would have liked to make the ball look more bouncy, with appropriate dents and a proper texture map. A basket ball court would have been nice too. Note that you can draw on the animation as it plays in different colors and line widths. I would have liked to be able to save the edits in a file with the animation frames modified, but this turned out to be difficult. Also, the picture shows a coherent view of the desktop, with three frames that could be placed where ever the user wants. This makes the animation frame more minimalist and thus suggests a movie screen. One can make a case for combining the other two frames, however. Sorry about the image quality.
Project Wrapup and Future Directions
This was an individual project. I used Java to construct the GUI and perform the animation, and BMRT to create all of the non-trivial animation images, including the icon animations. I was able to create an interface for interacting with animations using gestures. Two main features of using gesture-based interaction are providing speed information during playback and the ability to edit the animation. The editing I've accomplished is very simple. I wrote the skeleton for changing the editing attributes but only color changes are now supported. I created a total of 8 animation sequences, 6 of which are intended for gesture explanations (see section above). The tool bar is the most complete part of the project. It should be intuitive for everyone to use (I hope).
I've done more work on the interface than I had expected. For example, I chose to implement the animation of the button icons, which took a lot of time. I did not write sophisticated shaders or used complex lighting models. I had intended to fully utilize the shading language in BMRT. Due to time constraints, I ended up using only the basics of RenderMan. The modelling, however, was appropriate for the scenes I constructed.
In the future, I hope to extend Penimation to allow for different ways of editing the animation sequence. There are two main ways to go here. One can edit the animation as it plays, creating a "trail" that can be displayed when the sequence is displayed again. One can also edit individual frames which are extrapolated to multiple frame changes using some interpolation algorithm. Both of these topics would give Penimation the real edge over other animation players because this is where the real power of gestures come in. I had experimented with gestures in this project by using them to determine play speed and adding strokes to a frame. I need to, however, take the pen-animation interaction idea further by daring to accomplish everything you could do with a traditional animation interface using a gesture-based interface.
Some things I can do in the near future follow. I can play the command button animation once everytime the corresponding gesture is made on the main animation frame. This gives feedback that the gesture was interpreted. I also want to implement a progress bar for loading long animation sequences. It is also easy to make the animations themselves look a lot nicer, using RenderMan's texture map, bump map, and other utilities. Finally, I want to extend the editing support to allow for typing in text on the animations, and inserting images. The difficult part there is to save the pictures to some archive somewhere. I need to figure out how to do that without having to mess around with the internals of the file compression format.
Created by Ray Luo
|up | top  |