Synthetic Image Detection Challenge Call for Papers
We are pleased to invite you to participate in the second edition of the
Synthetic Image Detection Challenge,
organized as part of the MediaEval 2026 workshop.
Building on the strong participation and insights of the first
edition, this year's challenge continues to focus on the detection
of synthetic visual content and the localization of manipulated
regions in images collected from real-world, in-the-wild sources.
The challenge consists of two complementary tasks:
Task A: Real vs. Synthetic Image Classification
Task B: Manipulated Region Localization in AI-modified Images
What's new in 2026:
Task A remains unchanged, ensuring continuity and comparability
with last year's results.
Task B introduces a new test set, reflecting more recent
manipulation patterns and generative content characteristics.
Stronger emphasis beyond rankings and metrics: this edition
explicitly encourages contributions that provide deeper insights
into the evolving landscape of generative AI and its implications
for media forensics. Understanding how new generative models,
workflows, and real-world transformations shape synthetic traces
is of high value to the community. Participants are therefore
encouraged to share analyses, ablation studies, and qualitative or
quantitative observations that help advance our collective
understanding of this rapidly changing domain.
We welcome participation from researchers and practitioners working in
computer vision, multimedia forensics, machine learning, and related
fields.
More details about the challenge — including task descriptions,
evaluation protocols, and important deadlines — are available
here: https://multimediaeval.github.io/editions/2026/tasks/synthim/
To register and join the challenge, please visit the MediaEval 2026
website: https://multimediaeval.github.io/editions/2026/
The dataset and submission instructions can be found in the official
GitHub
repository: https://github.com/mever-team/mediaeval2026-sid
We look forward to your participation and to seeing your contributions
at MediaEval 2026.
Best regards,
Olga Papadopoulou
on behalf of the Synthetic Image Detection Task Organizers