Title: | Depth Completion for Close-Range Specular Objects |
Authors: | Pourmand, S. Merillou, N. Merillou, S. |
Citation: | WSCG 2022: full papers proceedings: 30. International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision, p. 135-141. |
Issue Date: | 2022 |
Publisher: | Václav Skala - UNION Agency |
Document type: | conferenceObject |
URI: | http://hdl.handle.net/11025/49587 |
ISBN: | 978-80-86943-33-6 |
ISSN: | 2464-4617 |
Keywords: | dokončení hloubky;RGB-D obrázky;syntetický datový soubor;zrcadlové odrazy |
Keywords in different language: | depth completion;RGB-D images;synthetic dataset;specular reflections |
Abstract in different language: | Many objects in the real world exhibit specular reflections. Due to the limitations of the basic RGB-D cameras, it is particularly challenging to accurately capture their 3D shapes. In this work, we present an approach to correct the depth of close-range specular objects using convolutional neural networks. We first generate a synthetic dataset containing such close-range objects. We then train a deep convolutional network to estimate normal and boundary maps from a single image.With these results, we propose an algorithm to detect the incorrect area of the raw depth map. After removing the erroneous zone, we complete the depth channel. |
Rights: | © Václav Skala - UNION Agency |
Appears in Collections: | WSCG 2022: Full Papers Proceedings |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
B89-full.pdf | Plný text | 3,59 MB | Adobe PDF | View/Open |
Please use this identifier to cite or link to this item:
http://hdl.handle.net/11025/49587
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.