US 12,220,640 B2
Generative interior design in video games
Han Liu, Millbrae, CA (US); Jingwen Liang, Mountain View, CA (US); Schaefer Mitchell, Beaverton, OR (US); and Mohsen Sardari, Burlingame, CA (US)
Assigned to ELECTRONIC ARTS INC., Redwood City, CA (US)
Filed by Electronic Arts Inc., Redwood City, CA (US)
Filed on Sep. 1, 2021, as Appl. No. 17/463,754.
Prior Publication US 2023/0061250 A1, Mar. 2, 2023
Int. Cl. A63F 13/63 (2014.01); G06F 30/12 (2020.01); G06F 30/13 (2020.01); G06N 3/049 (2023.01)
CPC A63F 13/63 (2014.09) [G06F 30/12 (2020.01); G06F 30/13 (2020.01); G06N 3/049 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented generative interior design method to customize an interior region of a video game building in accordance with a user's preferences while playing the video game, comprising:
obtaining, based on user input directed to a computing device of a user, input data comprising:
(i) boundary data defining a boundary of the interior region of the video game building, and
(ii) constraint data defining one or more constraints for the interior region, wherein the constraint data comprises a budget that defines a maximum of in-game currency for the user to spend on objects to be placed in the interior region;
generating a floor plan for the interior region of the video game building, wherein the floor plan divides the interior region into a plurality of interior spaces, comprising processing the input data using a floor plan generator model;
generating, by a layout generator model comprising one or more graph neural networks, a layout for at least one of the plurality of interior spaces defined by the floor plan, the layout representing a configuration of one or more objects to be placed in the interior region;
wherein generating the layout comprises:
initializing an instance of a graph to be operated on by a graph neural network of the one or more graph neural networks;
for each iteration of a plurality of iterations:
adding a node to the graph, wherein each node of the graph represents an object to be placed in the interior region; and wherein adding a node to the graph comprises processing a current representation of the graph using the graph neural network comprising:
generating a probability distribution over node types;
selecting a node type based upon the generated probability distribution over node types; and
generating the node for adding to the graph based upon the selected node type; and
generating, by the graph neural network, edges between nodes of the graph, wherein an edge between two nodes specifies relational information between the objects corresponding to the nodes; and
causing the interior region to be visually rendered for presentation to the user via a display of the computing device during gameplay of the video game and in response to additional user input.