Although not directly applied, this is similar to how DLSS 3.0 Frame Generation generates frames in between GPU-rendered frames to increase the frame rate in games. This is done by taking 2 rendered frames and appending an AI-created frame in between both frames that were created from extracting the motion vectors of the pixels on the screen. This is similar to how keyframes use starting and ending points to create motion, but this case is done in frame intervals by extracting motion vectors + using a start and end frame.
More detail can be found here:
https://www.nvidia.com/en-us/geforce/news/dlss3-ai-powered-neural-graphics-innovations/
michelllepan
I recently read an interesting human-robot interaction paper that takes inspiration from the idea of keyframes in animation to introduce a new method for people to teach robots new skills. The main idea is to allow humans to manipulate a robot arm to capture keyframes along a motion they want to demonstrate rather than having to demonstrate an entire trajectory, which is similar to the goal of reducing the work the animator must do in creating a scene. (https://www.cc.gatech.edu/social-machines/papers/akgun12_hri_keyframes.pdf)
It's cool to see the many connections between animation and robotics :)
Although not directly applied, this is similar to how DLSS 3.0 Frame Generation generates frames in between GPU-rendered frames to increase the frame rate in games. This is done by taking 2 rendered frames and appending an AI-created frame in between both frames that were created from extracting the motion vectors of the pixels on the screen. This is similar to how keyframes use starting and ending points to create motion, but this case is done in frame intervals by extracting motion vectors + using a start and end frame.
More detail can be found here: https://www.nvidia.com/en-us/geforce/news/dlss3-ai-powered-neural-graphics-innovations/
I recently read an interesting human-robot interaction paper that takes inspiration from the idea of keyframes in animation to introduce a new method for people to teach robots new skills. The main idea is to allow humans to manipulate a robot arm to capture keyframes along a motion they want to demonstrate rather than having to demonstrate an entire trajectory, which is similar to the goal of reducing the work the animator must do in creating a scene. (https://www.cc.gatech.edu/social-machines/papers/akgun12_hri_keyframes.pdf)
It's cool to see the many connections between animation and robotics :)