Press "Enter" to skip to content

Making computer animation more agile, acrobatic — and realistic

Animation in film and video games is hard to make realistic: each action typically requires creating a separate controller, while deep reinforcement learning has yet to generate realistic human or animal motion. Computer scientists have now developed an algorithm that uses reinforcement learning to generate realistic simulations that can even recover realistically, after tripping, for example. The same algorithm works for 25 acrobatic and dance tricks, with one month of learning per skill.

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *