I’m part of a group of Stanford students who are looking to build a way to guess/predict the location of a player’s body based off the head/hands locational data from the controllers. Do you all think a machine learning algorithm that takes a body of motion-capture data and predicts how the player’s arms, torso, and legs are moving would be useful for game development? What body parts would be most important, and what types of games do you think could utilize this technology best?
In jeneral it should be possible to olve this problem with IK methods I think, nut machine learning is fancy now, maybe it would be of use. Interesting to see the results.
All parts are important I think. Ok, all parts of upper torso. Legs can be only barely predicted, until we will have legs controllers.
Games? VR Boxing comes in mind.
Recently, I’m doing a similar project. To use a machine learning algorithm to predict head movement. Which algorithm do you think is suitable for this prediction?
Microsoft Kinect uses this sort of information but unfortunately I really don’t recall that many great games - mainly sport and dancing.
Could be very useful. But you probably want some ‘calibration’ data for the 3-point tracking conditions.
i.e. the NN algorithm has the 3 points of motion to work from… but also has data about overall height, shoulder height and width, hand length, waist height, etc that it might be able to better work from.
Have you seen this…
https://developer.oculus.com/blog/developer-perspectives-character-animation-in-dead-and-buried/
Go forth and do it. I am planning to dive into Machine Learning and UE4 on a later stage for a project and would like to talk to you guys. But I think even if the project is a failure you will definitely discover something most UE4 users never even got to dream of. You have to remember that we are all influenced by games and movies (mostly)…but if you read how machine learning and ai are just compiling things we as humans never even dreamed of