US 11,941,831 B2
Depth estimation
Tristan William Laidlow, London (GB); Jan Czarnowski, Gdanski (PL); and Stefan Leutenegger, Munich (DE)
Assigned to Imperial College Innovations Limited, London (GB)
Filed by Imperial College Innovations Limited, London (GB)
Filed on Jul. 23, 2021, as Appl. No. 17/384,359.
Application 17/384,359 is a continuation of application No. PCT/GB2020/050084, filed on Jan. 15, 2020.
Claims priority of application No. 1901007 (GB), filed on Jan. 24, 2019.
Prior Publication US 2021/0350560 A1, Nov. 11, 2021
Int. Cl. G06T 7/55 (2017.01); G05D 1/00 (2006.01); G06N 3/04 (2023.01); G06T 1/00 (2006.01); G06T 7/70 (2017.01)
CPC G06T 7/55 (2017.01) [G05D 1/0251 (2013.01); G06N 3/04 (2013.01); G06T 1/0014 (2013.01); G06T 7/70 (2017.01); G06T 2207/10016 (2013.01); G06T 2207/20076 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20221 (2013.01); G06T 2207/30244 (2013.01)] 32 Claims
OG exemplary drawing
 
1. An image processing system to estimate depth for a scene captured in a first frame of video data, comprising:
a fusion engine to receive a first depth estimate from a geometric reconstruction engine and a second depth estimate for the first frame of video data from a neural network architecture, and to probabilistically fuse the first depth estimate and the second depth estimate to output a fused depth estimate for the scene,
wherein the fusion engine is configured to receive a measurement of uncertainty for the first depth estimate from the geometric reconstruction engine and a measurement of uncertainty for the second depth estimate from the neural network architecture,
wherein the fusion engine is configured to use the measurements of uncertainty to probabilistically fuse the first depth estimate and the second depth estimate,
wherein the first depth estimate comprises a plurality of first depth estimates for the first frame of video data, at least one of the plurality of first depth estimates being generated using a second frame of video data that differs from the first frame of video data, and
wherein the fusion engine is configured to iteratively output the fused depth estimate for the scene, at each iteration processing the second depth estimate and one of the plurality of depth estimates.