US 12,307,733 B2
Learning illumination from diverse portraits
Chloe LeGendre, Culver City, CA (US); Paul Debevec, Culver City, CA (US); Wan-Chun Ma, Culver City, CA (US); Rohit Pandey, Mountain View, CA (US); Sean Ryan Francesco Fanello, San Francisco, CA (US); and Christina Tong, Pleasanton, CA (US)
Assigned to Google LLC, Mountain View, CA (US)
Appl. No. 17/309,171
Filed by Google LLC, Mountain View, CA (US)
PCT Filed Sep. 21, 2020, PCT No. PCT/US2020/070558
§ 371(c)(1), (2) Date May 3, 2021,
PCT Pub. No. WO2021/236175, PCT Pub. Date Nov. 25, 2021.
Claims priority of provisional application 62/704,657, filed on May 20, 2020.
Prior Publication US 2022/0027659 A1, Jan. 27, 2022
Int. Cl. G06V 10/60 (2022.01); G06F 18/214 (2023.01); H04N 23/611 (2023.01); H04N 23/741 (2023.01)
CPC G06V 10/60 (2022.01) [G06F 18/214 (2023.01); H04N 23/611 (2023.01); H04N 23/741 (2023.01)] 21 Claims
OG exemplary drawing
 
1. A method, comprising:
receiving an image representation of a first human; and
generating a prediction engine based on the image representation of the first human including producing a rendered image of a reference object and generating a comparison between the rendered image of the reference object and a ground truth image of the rendered image of the reference object, the comparison being based on a weight associated with the reference object, the prediction engine being configured to produce a predicted illumination profile based on an image representation of a second human.