Recent advances in generative AI have fundamentally transformed the visual ecosystem. Synthetic images, videos, avatars, and AI-generated content are now pervasive and increasingly integrated into domains such as online communication, biometric authentication, media creation, and human-computer interaction. While prior research has largely focused on the generation, manipulation, and detection of synthetic media, a broader challenge is emerging: ensuring the robustness, reliability, and trustworthiness of computer vision systems operating in environments increasingly dominated by synthetic content.
The growing prevalence of synthetic media is reshaping the data landscape for modern vision systems, introducing new forms of distribution shift, data contamination, and uncertainty in training and evaluation pipelines. These challenges are particularly relevant for large-scale and foundation vision models that rely on massive amounts of multimodal data and synthetic augmentation. Understanding how vision systems generalize, adapt, and fail under synthetic conditions is becoming a critical research problem.
This workshop aims to bring together researchers and practitioners working on robust vision systems in generative settings. We seek contributions spanning robustness, generalization, trustworthy evaluation, synthetic data governance, and deployment challenges across vision models, biometric systems, multimodal architectures, and foundation models. The workshop encourages interdisciplinary perspectives connecting computer vision, biometrics, machine learning robustness, AI safety, and media authenticity.
By shifting the focus from isolated detection problems toward robust visual intelligence in synthetic ecosystems, the workshop aims to define emerging research directions for trustworthy computer vision in the age of synthetic media.
Scope
We invite high-quality submissions that investigate the robustness, reliability, and trustworthiness of vision systems in the presence of synthetic media and generative AI. The workshop welcomes theoretical, empirical, and applied contributions that examine how vision systems behave, generalize, and fail under real-synthetic data mixtures, synthetic distribution shifts, and generative deployment settings. Topics of interest include, but are not limited to:
-
Robust vision under synthetic media:
- Distribution shift, failure modes, and robustness of vision systems in the presence of synthetic content.
-
Biometrics and identity systems:
- Robust biometric recognition, presentation attacks, training, and identity consistency in generative settings.
-
Foundation and multimodal models:
- Impact of synthetic data on large-scale training, pretraining/fine-tuning, and evaluation of foundation models.
-
Synthetic data and learning systems:
- Synthetic data generation, dataset contamination, data curation, and learning under mixed real-synthetic distributions.
-
Trustworthy and reliable vision systems:
- Calibration, uncertainty estimation, robustness evaluation, explainability, and failure analysis.
-
Vision for security and digital identity:
- Applications in authentication, digital identity, surveillance, and content integrity.
-
Deployment and real-world robustness:
- System-level robustness, operational challenges, and safety in real-world vision deployments.
-
Broader impacts of synthetic media:
- AI safety, societal implications, and content authenticity in vision pipelines.
Important Dates
RV-SE Workshop will follow a single-round, double-blind review process.
- Full Paper and Supplementary Submission: July 31, 2026
- Acceptance Notice: August 24, 2026
- Camera-Ready Submission: TBD
* All deadlines are based on the PDT time zone.
Speakers
Speakers will be announced soon. Please keep checking this website for updates.
Submission Guidelines
- Authors are invited to submit original, unpublished research papers related to the themes of the RV-SE workshop.
- All submissions must follow the BMVC 2026 paper formatting guidelines . Detailed formatting instructions and templates will be made available through the BMVC website.
- Submissions will be handled through the workshop submission portal. The submission link will be announced on this website.
- All papers will undergo a double-blind peer-review process. Authors must ensure that submissions are anonymized and comply with the BMVC reviewing policy.
- Paper Length: Submissions should follow the page limits specified by BMVC for workshop papers. Supplementary material may be submitted where permitted.
- At least one author of each accepted paper must register for BMVC 2026 and present the work at the workshop.
- Accepted papers will be presented as oral presentations and/or posters, depending on the workshop schedule and the number of accepted submissions.
- Authors are encouraged to release code, models, and datasets whenever possible to promote reproducibility and facilitate future research.
- Further details regarding submission procedures, important dates, and presentation instructions will be announced on the workshop website.
