Back to results
US 12,770,036 B2Granted
Artificial intelligence annotation through gamification
OwnerWesley T. Chuanginventor
Life of this patent
Estimating the term…
Abstract
In a system and method of labeling or annotating entities or objects in a data stream, the data is displayed in a virtual game environment, and the identification and labeling tasks are carried out by players of the game through game input devices capable of selecting an object or entity displayed in the game environment, and of categorizing the object or entity upon selection.
Claims17
A system for training an artificial intelligence (AI) to recognize virtual reality objects by enticing users to label the virtual reality objects, comprising:
- a server connected to a network and configured to broadcast a data stream over the network to a plurality of computing devices, the data stream including, in a virtual game environment, objects or entities recognizable by users of the computing devices, wherein the computing devices are configured to:
- display, on display devices connected to respective said computing devices, the objects or entities included in the data stream,
- enable users, in response to the display of one of the objects or entities included in the data stream, to play video games by manipulating or utilizing computer input devices to select and label, in a manner consistent with the virtual game environment, the displayed objects or entities,
- communicate, via the network, labels selected by the user for the displayed object or entity from the computing devices to the AI for use in training the AI and increasing model accuracy by identifying the object or entity based on labels input by multiple said users through the plurality of computing devices;
- upon selection of a label by a respective user, display a result, consistent with the virtual game environment, that rewards the respective user for carrying out the selection and labeling in order to entice the user into continuing to select and label displayed objects, and thereby train the AI to recognize the displayed objects or entities included in said data stream and broadcast over the network.
2 more elements
Dependencies are inferred from the claim text.