Special Issue "Learning 3D Reconstruction Without Supervision"

A special issue of Journal of Imaging (ISSN 2313-433X).

Deadline for manuscript submissions: closed (31 May 2021).

Special Issue Editor

Dr. Seungryong Kim
E-Mail Website
Guest Editor
Department of Computer Science and Engineering, Korea University, Seoul, Korea
Interests: computer vision; computational photography, machine learning, deep learning; representation learning; visual scene reconstruction; visual scene understanding

Special Issue Information

Dear Colleagues,

Perceiving the 3D geometric configuration of scenes or objects in an image has been essential for numerous tasks in computer vision, image processing, computational photography, and robotics applications, such as autonomous driving vehicles or drones, mobile robot localization and mapping, obstacle avoidance and path planning, and human behavior understanding. Due to the inherent challenge of this task in that an image is a partial measurement of scenes in 3D world, the 3D reconstruction is heavily under-constrained and thus has been notoriously considered as an unsolved problem. For the last few years, the communities have made revolutionary progress in precisely reconstructing 3D information from images with the advent of deep neural networks (DNNs), ranging from stereo matching to monocular depth estimation. Generally, they are largely based on a supervised learning paradigm requiring a huge amount of ground truth depth labels. However, achieving such labels is cumbersome and time-consuming, since particular hardware such as time-of-flight (ToF) and LiDAR or constrained multi-view camera settings are required, which hinders its applicability.

This Special Issue aims to answer the following question: Can we train the networks for 3D reconstruction without such training data? We request contributions presenting techniques that reconstruct 3D information with deep networks in a weakly, self-supervised, or unsupervised fashion. In particular, they may contain approaches for stereo matching, multi-view stereo matching, monocular depth estimation, or monocular human shape and pose estimation that can be trained without any 3D supervisions.

Dr. Seungryong Kim
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All papers will be peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Journal of Imaging is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 1600 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • 3D reconstruction
  • monocular depth estimation
  • stereo matching
  • multi-view stereo matching
  • monocular human shape and pose estimation

Published Papers

There is no accepted submissions to this special issue at this moment.
Back to TopTop