We have been able to complete most of the preliminary framework of the project
Subhadip has implemented the representation for Motion States by
creating data structures for pose and an associated velocity. He has
also setup the framework for the motion database which is a collection
of motion states.
Leslie has implemented the similarity weights between the motion
states and is currently making progress on integrating the approximate nearest
neighbor (ANN) implementation of the k-NN algorithm to work with our
data structures
Sean has continued working on the Maya/ MEL/ C++ interface and looking
in to methods of adding interactive control to through Maya. He also
updated the Maya plug-in to use the Maya joint structures, and
splicing and interpolation work with the new structures.
Sajika is working on the reinforcement learning implementation and
researching ways integrate with motion data. He is implementing a
basic Q-function and setting it up to work with the motion states.
Maya screenshot of interpolating from ballet to breakdancing