Half-century of Structure-from-Motion
50SFM - ECCV '24 Workshop
Allianz MiCo - Milano Convention Centre, Milan, Italy
29 September 2024, 14:00 -18:00, Space 2
ABOUT
The 50SFM workshop wants to commemorates the remarkable journey of Structure-from-Motion (SFM) over the past 50 years. SFM has played a pivotal role in the fields of computer vision, photogrammetry and robotics, revolutionizing our ability to reconstruct 3D structures from 2D image sequences.
Originally coined in neuroscience in reference to the human visual system, it was Ullmann in 1977 (ftp://publications.ai.mit.edu/ai-publications/pdf/AIM-476.pdf) who first applied it within a computational context, albeit at the intersection of neuroscience. In his seminal work in 1977 - published also in 1979 by MIT, the interpretation of structure from motion is examined from a computational point of view. "The question addressed is how the 3-D structure and motion of objects can be inferred from the 2-D transformations of their projected images when no 3-D information is conveyed by the individual projections". He articulated the so-called SFM theorem: given three orthographic views of four non-coplanar points, the structure and motion compatible with these views are uniquely determined.
This theorem sparked interest in the computational structure from motion problem, reaching its peak in the early 1990s, gradually declining later on until it experienced a resurgence after 20120 (see graph below from Google Books NGrams).
This ECCV 2024 workshop aims to bring together experts, researchers and enthusiasts to reflect on the evolution of SFM, discuss recent advances and explore its diverse applications. The main goal is to provide a venue for knowledge exchange, collaboration and future directions in the field of SFM. With a mix of invited speakers and solicit papers, the workshop aims to:
● Provide a comprehensive overview of the historical development of SFM;
● Explore recent breakthroughs, theoretical results and innovations in SFM algorithms and methodologies;
● Discuss challenges and potential solutions in the current state of SFM research;
● Showcase diverse applications of SFM across industries, including robotics, archaeology, geosciences, and more;
● Foster networking and collaboration among researchers and practitioners in the SFM community.
The expected outcomes of the 50SFM workshop are:
Increased awareness of the historical context and evolution of SFM;
Enhanced understanding of recent advancements in SFM algorithms and methodologies;
Identification of challenges and potential solutions in the current state of SFM research;
Establishment of collaborations and networking opportunities within the SFM community;
Inspiration for future research directions and applications of SFM.
The workshop is relevant to the computer vision, photogrammetry and robotics communities as it serves as a platform for knowledge exchange, collaboration, and inspiration. By addressing historical context, current challenges, and future directions in SFM, the workshop contributes to the continuous advancement of this pivotal technology and strengthens the bonds within the computer vision research community.
The 50SFM workshop is relevant to the computer vision, photogrammetry and robotics communities as it serves as a platform for knowledge exchange, collaboration, and inspiration. By addressing historical context, current challenges, and future directions in SFM, the workshop contributes to the continuous advancement of this pivotal technology and strengthens the bonds within the computer vision research community.
50SFM TOPICS and THEMES
Historical review of SFM: A journey through the key milestones and developments in SFM over the past years
State-of-the-Art of SFM Algorithms: Presentations and discussions on the latest advancements in SFM methodologies and techniques, including learning-based matching methods, graph-based solutions, incremental vs global methods, etc.
Applications of SFM: Case studies and presentations highlighting the diverse applications of SFM in various domains
Challenges in SFM: Identifying and addressing current challenges in SFM research and implementation
Influence of deep learning in modern SfM solutions
SfM, Visual Odometry and Visual SLAM: SfM and its “sisters” for real-time applications
SfM and geolocation: Integration of SfM outcomes with GNSS data
Visual localization in the wild: Visual place recognition for loop closing in Visual SLAM
Metric accuracy of SfM for large scene reconstruction
Future Directions: Panel discussions and talks on the future of SFM, emerging trends, and potential breakthroughs
CHALLENGING SEQUENCES
Organizers collected various datasets which are challenging for SfM approaches. They include repeated structures, transparent objects, reflective surfaces, etc. Please contact us if you wish to receive the sequences.
Participants are welcome to use these sequences to test their methods and report solutions and results at the event and in conference papers. Submitted papers will be reviewed and, if accepted, published in the ECCV 2024 proceedings.
Sequence 1 - repeated structures (Temple)
Sequence 2 - repeated structures (Street View)
Sequence 3 - repeated structures (Drone)
Sequence 4 - transparent objects (Glass)
Sequence 5 - Forestry
INVITED SPEAKERS
Czech Technical University, Czech Republic
"How to move the goalposts: coming up with yet unsolved challenges for SfM"
ORGANIZERS
PAPER SUBMISSIONS & PROCEEDINGS
Jul 19th, 2024: Paper submission
Aug 9th, 2024: Notification to authors
Aug 25th, 2024: Camera ready for inclusion in ECCV proceedings
All submissions will be handled electronically via the CMT conference submission website: https://cmt3.research.microsoft.com/50SFM2024
All authors must agree to the policies stipulated below. Papers must be registered in CMT by the paper registration deadline.
All papers will be reviewed by at least two reviewers with double-blind peer-review policy.
TECHNICAL PROGRAM
The 50SFM workshop will be conducted on September 29th in the afternoon in room SPACE 2 and it will consist of a series of invited presentations, pitch talks from the submitted papers and poster presentations.
The schedule of the workshop and the accepted papers are listed below:
14:00 - 14:05: Welcome & Intro - 5SFM organizers
14:05 - 14:30: Spotlight / pitch presentations of accepted papers
D. Di Nucci et al. - KRONC: Keypoint-based Robust Camera Optimization for 3D Car Reconstruction
S. Hun Lee et al. - What's Wrong with the Absolute Trajectory Error?
S. Paul et al. - MPVO: Motion-prior based Visual Odometry for Pointgoal Navigation
C. Tzamos et al. - Are Minimal Radial Distortion Solvers Necessary for Relative Pose Estimation?
X. Wang et al. - LVG-SfM: Learning-based View-Graph Generation for Robust On-the-Fly SfM
14:30 - 15:00: "A historical review of SfM - 300 years and counting"
P. Sturm (INRIA, France) - Invited talk
15:00 - 15:30: "How to move the goalposts: coming up with yet unsolved challenges for SfM"
D. Mishkin (Czech Technical University, Czech Republic) - Invited talk
15:30 - 16:00: Coffee break & poster presentations
16:00 - 16:30: "My 30-year journey with SfM"
M. Pollefeys (ETH Zurich / Microsoft, Switzerland) - Invited talk
16:30 - 17:00: "Real-time feature-based visual SLAM"
J.M. Martínez Montiel (Universidad de Zaragoza, Spain) - Invited talk
17:00 - 17:30: "Value creation by using SfM"
C. Strecha (Pix4D, Switzerland) - Invited talk
17:30 - 18:00: "Certifiable pose estimation and 3D reconstruction beyond geometry"
L. Carlone (MIT, USA) - Invited talk
18:00: Closing remarks
POLICIES
50SFM workshop follows the general ECCV 2024 submission policies. ECCV 2024 uses the Springer Nature Code of Conduct for Book Authors as the basis for many of the policies. We therefore urge you to familiarize yourself with these guidelines before preparing and submitting your work to ECCV 2024.
In submitting a manuscript to ECCV, the authors acknowledge that no paper substantially similar in content has been or will be submitted to another conference or workshop during the review period. Please refer to the Submission Policies on the conference web site for additional details on dual submissions and guidelines concerning prior work.
By submitting a paper to ECCV, the authors agree to the review process and understand that papers are processed by CMT, TPMS (Toronto Paper Matching System), as well as OpenReview to match each manuscript to the best possible area chairs and reviewers. Moreover, the authors agree that papers will be checked for plagiarism using iThenticate.
The authors should be aware that each accepted paper is expected to be presented at ECCV in-person by an author (or an authorized delegate).
All accepted papers will be made publicly available by Springer and/or the European Computer Vision Association (ECVA) no earlier than four weeks before the conference. Authors wishing to submit a patent understand that the paper’s official public disclosure is four weeks before the conference or whenever the authors make it publicly available, whichever is first. More information about ECVA is available at https://www.ecva.net/.