Please use this identifier to cite or link to this item: http://dx.doi.org/10.14279/depositonce-11179.2
For citation please use:
Main Title: Pixel2point: 3D object reconstruction from a single image using CNN and initial sphere
Author(s): Afifi, Ahmed J.
Magnusson, Jannes
Soomro, Toufique A.
Hellwich, Olaf
Type: Article
Language Code: en
Abstract: 3D reconstruction from a single image has many useful applications. However, it is a challenging and ill-posed problem as various candidates can be a solution for the reconstruction. In this paper, we propose a simple, yet powerful, CNN model that generates a point cloud of an object from a single image. 3D data can be represented in different ways. Point clouds have proven to be a common and simple representation. The proposed model was trained end-to-end on synthetic data with 3D supervision. It takes a single image of an object and generates a point cloud with a fixed number of points. An initial point cloud of a sphere shape is used to improve the generated point cloud. The proposed model was tested on synthetic and real data. Qualitative evaluations demonstrate that the proposed model is able to generate point clouds that are very close to the ground-truth. Also, the initial point cloud has improved the final results as it distributes the points on the object surface evenly. Furthermore, the proposed method outperforms the state-of-the-art in solving this problem quantitatively and qualitatively on synthetic and real images. The proposed model illustrates an outstanding generalization to the new and unseen images and scenes.
URI: https://depositonce.tu-berlin.de/handle/11303/12336.2
http://dx.doi.org/10.14279/depositonce-11179.2
Issue Date: 23-Dec-2020
Date Available: 8-Jan-2021
DDC Class: 004 Datenverarbeitung; Informatik
Subject(s): single-view reconstruction
deep learning
point cloud
CNN
Sponsor/Funder: TU Berlin, Open-Access-Mittel – 2020
License: https://creativecommons.org/licenses/by/4.0/
Journal Title: IEEE Access
Publisher: IEEE
Publisher Place: New York, NY
Volume: 9
Publisher DOI: 10.1109/ACCESS.2020.3046951
Page Start: 110
Page End: 121
EISSN: 2169-3536
Appears in Collections:FG Computer Vision & Remote Sensing » Publications

Files in This Item:

Version History
Version Item Date Summary
2 10.14279/depositonce-11179.2 2021-01-08 10:25:16.235 final published version
1 10.14279/depositonce-11179 2021-01-04 15:17:33.0
Item Export Bar

This item is licensed under a Creative Commons License Creative Commons