US 12,154,212 B2
Generating environmental data
Zhenpei Yang, Austin, TX (US); Yuning Chai, San Mateo, CA (US); Yin Zhou, San Jose, CA (US); Pei Sun, Palo Alto, CA (US); Henrik Kretzschmar, Mountain View, CA (US); Sean Rafferty, Sunnyvale, CA (US); Dumitru Erhan, San Francisco, CA (US); and Dragomir Anguelov, San Francisco, CA (US)
Assigned to Waymo LLC, Mountain View, CA (US)
Filed by Waymo LLC, Mountain View, CA (US)
Filed on Nov. 16, 2020, as Appl. No. 17/098,943.
Claims priority of provisional application 62/936,326, filed on Nov. 15, 2019.
Prior Publication US 2021/0150799 A1, May 20, 2021
Int. Cl. G06T 17/05 (2011.01); B60W 30/095 (2012.01); B60W 50/06 (2006.01); B60W 60/00 (2020.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06T 7/70 (2017.01); G06T 9/00 (2006.01); G06T 15/00 (2011.01); G06T 15/04 (2011.01)
CPC G06T 15/04 (2013.01) [B60W 30/0956 (2013.01); B60W 50/06 (2013.01); B60W 60/001 (2020.02); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06T 7/70 (2017.01); G06T 9/002 (2013.01); G06T 15/00 (2013.01); G06T 17/05 (2013.01); B60W 2420/403 (2013.01); B60W 2420/408 (2024.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30248 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
performing a scene reconstruction process comprising:
obtaining a surfel map generated from sensor observations of a real-world environment, wherein the surfel map assigns a surfel to each of a plurality of points in a three-dimensional space of the real-world environment, and
generating, for each surfel in the surfel map, a surfel textures; and
performing a synthesis process to generate simulated sensor data from a simulated location in the surfel map, the synthesis process comprising:
generating a textured surfel rendering from surfel textures visible from the simulated location,
generating one or more segmentation maps from the surfel map, and
processing an input comprising: (i) the textured surfel rendering generated from the surfel textures, and (ii) the one or more segmentations maps using a generative adversarial network to generate the simulated sensor data from the simulated location.