MIT researchers have developed a technique that could revolutionize the animation industry by giving artists more flexibility in how they animate characters. Instead of sticking to a single conventional measurements or mathematical functions called barycentric coordinates, the new method allows the artist to experiment with different movements and expressions, specific to each individual animation. This innovation opens up room for creativity and individual artistry in creating animations that more closely align with the artist’s vision. Furthermore, this technique could find applications in other fields such as medical imaging, architecture, and virtual reality.
Unlike conventional animation techniques that provide only a set of fixed barycentric coordinate functions, this new method seeks to make the entire process flexible and artist-friendly. The flexibility lies in allowing the artist to design the look and feel of animations by choosing among diverse and flexible smoothness energies generated by the function, all of which ultimately influence the movement and appearance of the animation.
The concept of barycentric coordinates stemmed from the work of German mathematician August Möbius in 1827. However, this old mathematical construct proves challenging when it comes to designing complex shapes beyond simple triangles. As a solution to this complexity, the MIT research team used a type of neural network to model the coordinates. Neural networks, given their ability to process inputs through layers of connected nodes, are perfect to generate valid solutions that respect all constraints.
This creative process involves layering the shape with overlapping virtual triangles, which are then combined through the neural network to produce a smoothly functioning, yet complex character movement. The researcher’s network architecture is specifically designed to satisfy all constraints. This technique allows artists to experiment with different motions until they reach the perfect animation.
The expanded flexibility provided by the neural network is one of the most significant aspects of this method. Beyond accurately modelling the look and feel for animations, the technique can be used to create more accurate, life-like movements that are even more engaging and realistic.
Future plans for this research include working to accelerate the neural network and creating an interactive interface to allow artists to make real-time changes to their animations. This groundbreaking work was recognized and projected at SIGGRAPH Asia and funded by numerous significant organizations. The successful application of this research promises to bring a tangible difference to the world of animation, pushing the boundaries of artistic freedom and creative expression.
