SCIFOR WORKSHOP
SCIENTIFIC IMAGE FORENSICS WORKSHOP
ICIP 2026
13-17 SEPTEMBER 2026, TAMPERE, FINLAND
SCIFOR WORKSHOP
SCIENTIFIC IMAGE FORENSICS WORKSHOP
ICIP 2026
13-17 SEPTEMBER 2026, TAMPERE, FINLAND
The number of scientific paper retractions due to forged images has quadrupled over the past two decades, mainly due to image duplication and manipulation. Image duplication, particularly copy–move forgery, remains a primary problem. Yet, detection in biomedical figures remains largely dependent on manual screening, as automatic general-purpose forensic tools often fail in the complex domain of biomedical imaging.
To catalyze solutions to the problem, we organized the RECOD.ai/LUC Scientific Image Forgery Detection Competition, sponsored by the IEEE Signal Processing Society and Kaggle research programs. Featuring a novel benchmark dataset derived from over 2,000 retracted articles, the competition attracted more than 1,500 teams worldwide, yielding a diverse set of approaches and practical insights into what works and fails when cutting-edge AI solutions are deployed to scientific images.
This workshop will present the findings from this competition, detailing the dataset, evaluation protocols, and the practical lessons learned when deploying AI for research integrity. Winning teams are invited to present their solutions and design choices, providing a snapshot of the current state of the art. The workshop will also feature contributed papers on image forensics applied to scientific images, and it will conclude with a panel of experts from image forensics and biomedical imaging to discuss emerging threats, key challenges, and open questions for the community.
Submitted manuscripts are invited on scientific image forgery, including (but not limited to) the following topics:
Scientific image forgery datasets and benchmarks;
Detection, localization, and instance-level segmentation of scientific forgeries (e.g., copy–move and AI-generated images);
Robustness and generalization across acquisition modalities and post-processing pipelines;
Forensics and counter-forensics: emerging attacks (including generative and editing pipelines) and practical defenses for scientific images;
Publisher and workflow-oriented methods for scalable screening, including human-in-the-loop review;
Ethical guidance and discussion on scientific image publishing
Submissions may include up to 5 pages + 1 page of only references
Submissions must follow the 2026 ICIP main track double-blind policy and formatting instructions: https://2026.ieeeicip.org/author-kit/
The review process will not include a rebuttal phase.
Submission webpage: https://icip2026.exordo.com/ (Satellite Workshop Papers Track)
Submission open date March 19, 2026
Submission due May 13, 2026
Notification of acceptance June 10, 2026
Camera-ready paper due July 1, 2026
Author registration due date: July 16, 2026
The workshop will be a half-day event with the following sections:
Recod.ai/LUC - Scientific Image Forgery Detection Challenge Overview and Lessons Learned
The organizers will introduce the scientific image forgery detection area, describe the competition benchmark, and present an analysis of the submitted challenge solutions in what worked and what didn’t across the competition.
Paper Presentations
Authors from accepted papers will present their work.
Expert Panel Section on Scientific Image Integrity.
A panel featuring two invited experts will discuss emerging threats, open research problems, and realistic paths to deployment in publisher and institutional workflows. The discussion will bring together perspectives from (a) digital forensics, (b) biomedical image analysis, and (c) research integrity, and will conclude with a moderated Q&A to connect the technical themes to actionable directions for the community.
The RECOD.ai/LUC Scientific Image Forgery Detection Competition was kindly supported by the Kaggle Research Grant Program and the IEEE Signal Processing Society Challenge Program.
We are grateful to Dr. Elisabeth Bik for her valuable guidance during the early phases of the project, and to members of the São Paulo Research Foundation (FAPESP) Horus project and the CNPq Aletheia research project for their technical support.
We extend our gratitude to Abigail Daman, Alessandra Vellucci Solari, Christina Truszkowski, and Emil Shahbazov for their help with constructing the competition dataset.