US 12,293,572 B2
Method and device for generating training data to generate synthetic real-world-like raw depth maps for the training of domain-specific models for logistics and manufacturing tasks
Yakov Miron, Haifa (IL); and Yoel Shapiro, Kiryat Bialik (IL)
Assigned to ROBERT BOSCH GMBH, Stuttgart (DE)
Filed by Robert Bosch GmbH, Stuttgart (DE)
Filed on Aug. 9, 2022, as Appl. No. 17/883,935.
Prior Publication US 2023/0055538 A1, Feb. 23, 2023
Int. Cl. G06V 10/774 (2022.01); G06V 10/82 (2022.01)
CPC G06V 10/774 (2022.01) [G06V 10/82 (2022.01)] 10 Claims
OG exemplary drawing
 
1. A computer-implemented method for providing training data for training of a data-driven model as a machine-learning model, wherein the data-driven model is to be trained to generate dense depth maps from sensor acquired raw depth maps, the method comprising the following steps:
providing multiple synthetic dense depth map data items from CAD data of various synthetic scenes;
providing multiple real raw depth map data items obtained from real-world depth sensor measurements of real-world scenes;
training a generative model to obtain a trained generator model for generating generated raw depth map data items from the synthetic dense depth map data items; and
applying the trained generator model to generate training data from provided synthetic dense depth map data.