Salient Poses converts hard-to-edit mocap into easy-to-edit keyframe animation. The idea is that you provide an animation, which Salient Poses analyzes to find potential sets of keyframes. You choose one set of keyframes that’s best for your use case, and then the Salient Poses creates a new animation for you. The animation is a bit like a blocked animation, but has the detail of a finished animation.
Before and after applying Salient Poses. The white dots show the data; one dot for each keyed in value. Notice that the “after” image has fewer dots while preserving the shape of each curve.
Why It’s Good
After applying Salient Poses, the simplified animation has fewer keyframes. This is great for compression, because less data is user to represent the animation. This is also great for editing, because the simplified animation can be adjusted using standard keyframe animation techniques. Almost all animation tools - such as Maya and Blender - provide support for adjusting motion using keyframes.
How It works
Given mocap, Salient Poses performs an optimal keyframe-reduction. The idea of the reduction is that we keep only the most important keyframes. Salient Poses provides not only one solution, but an entire range of solutions (the different solutions offer different levels of compression). After you choose the right solution for you, Salient Poses performs curve fitting to replace all non-keyframes with inbetweens that are configured to best recreate the original animation.
If you’re a technical artist, a hobbyists, or otherwise interested in using Salient Poses I’d recommend starting with the interactive Maya Plugin. Otherwise, developers looking to automate the process might be more interested in the Command Line Interface.
Otherwise, if you’re want to read more about the algorithm you might be interested to checkout the algorithm page for a more detailed overview of Salient Poses and also the results page to see some interactive examples.
Get in Touch
Beyond that, anyone is welcome to join our Slack Community and talk with me directly - any of your questions, feature requests, critique, or general chat are most welcome!
Richard Roberts developed Salient Poses during his doctorate study at Victoria University of Wellington. Taking inspiration from a SIGGRAPH sketch, Richard designed the algorithm as a new approach to compressing and editing motion capture animation.
Throughout the process, John Lewis supervised the project, helping out with both the high level theory and also the nuts of bolts are the programming (Lewis is now a lead researcher at SEED, Electronic Arts). The research team at OLM Digital - especially Ken Anjyo, who created and lead the team - contributed invaluable industry expertise and perspective to the research. Finally, Jaewoo Seo and Yeongho Seol reviewed the work many times over and contributed feedback from the side lines.
Outside of the everyone above, special thanks should be given to Ayumi Kimura and other staff at OLM Digital, along with artists and researchers at SEED, Electronic Arts, Weta Digital, and Victoria University of Wellington’s Computational Media Innovation Centre and Virtual World’s Lab. With the help of these people, the research would have been able to succeed.