Related Experiment Video
Updated: Aug 16, 2025
12:49
A Method for 3D Reconstruction and Virtual Reality Analysis of Glial and Neuronal Cells
Published on: September 28, 2019
12.9K
Iterative Online 3D Reconstruction from RGB Images
Thorsten Cardoen1, Sam Leroux1, Pieter Simoens1
1IDLab, Department of Information and Technology, Ghent University-imec, 9052 Ghent, Belgium.
Sensors (Basel, Switzerland)
|December 23, 2022
Summary
This study enhances online 3D reconstruction by modifying the Pix2Vox++ architecture for real-time applications. It also presents methods for selecting informative viewpoints to improve reconstruction quality and efficiency.
Area of Science:
- Computer Vision
- Robotics
Background:
- 3D reconstruction algorithms typically operate in offline settings, processing batches of images.
- Real-time applications like robot navigation require accurate 3D data during image acquisition.
Purpose of the Study:
- To adapt existing batch-based 3D reconstruction algorithms for online, iterative use.
- To improve the efficiency and accuracy of 3D reconstruction in dynamic environments.
Main Methods:
- Modification of the Pix2Vox++ architecture for online 3D reconstruction.
- Development and evaluation of viewpoint selection strategies for iterative reconstruction.
Main Results:
- Batch-based algorithms yield suboptimal results in online 3D reconstruction.
- Modified Pix2Vox++ and effective viewpoint selection achieve state-of-the-art reconstruction quality.
Conclusions:
- Online 3D reconstruction requires specialized algorithms and viewpoint selection.
- Efficient viewpoint selection is crucial for mitigating memory loss and computational load in real-time systems.
