Abstract
We present a learned method for structured-light depth estimation, where the projected pattern is jointly optimized alongside the depth-matching network. An in-network Tenderer simulates pattern projection during training. Results are precise, achieving 1/30th-pixel plane RMSE.
© 2020 The Author(s)
PDF Article | Presentation VideoMore Like This
Julie Buquet, Jinsong Zhang, Patrice Roulet, Jean-Francois Lalonde, and Simon Thibault
DF3A.1 3D Image Acquisition and Display: Technology, Perception and Applications (3D) 2020
Ki-Chul Kwon, Ki Hoon Kwon, Munkh-Uchral Erdenebat, Young-Tae Lim, Jong-Rae Jeong, Min Young Kim, and Nam Kim
P4_11 Conference on Lasers and Electro-Optics/Pacific Rim (CLEO/PR) 2020
Brian K. Spears
HTu3B.1 High Intensity Lasers and High Field Phenomena (HILAS) 2020