Atlas record: https://www.aitimeline.dev/en/events/evt-0218/
Availability · November 4, 2010
Kinect
Microsoft released a cheap depth camera that tracked a person's pose in real time with no controllers.
Why it matters
Three-dimensional perception cost a hundred and fifty dollars, and robotics laboratories bought it faster than gamers did.
Pose estimation was built on a randomised decision forest trained on hundreds of thousands of synthetic depth images of the body, running on the console's own graphics processor. The side effect outgrew the intent: Kinect became the standard cheap depth sensor in robotics and vision research, because laser scanners had cost far more. It is also one of the first mass-market products with machine learning running on the device itself.
Event record
- Event date
- November 4, 2010
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 22, 2026
- Lines
- ID
- evt-0218
The device was released on 4 November 2010.
Related events
Earlier
- Related The iPhone 2007
A second mass-market sensor, this time in the room rather than the pocket.
- Enables Random forests 2001links to this record
Kinect's body-part recognition is a randomized decision forest; the 2011 paper cites Breiman 2001 in its section on those forests.
- Related "Put-That-There": voice and gesture in one command 1980links to this record
Thematic: thirty years later Kinect sold voice and gesture as one input; no source read connects the two.