Computer vision has made remarkable progress on curated benchmarks, yet the transition to unconstrained real-world environments remains an open challenge. In practical deployments, vision systems face distribution shifts, dataset biases, adverse weather and illumination, sensor variability, domain changes, limited computational resources, and unpredictable operational scenarios, factors that can substantially degrade their performance and trustworthiness. WILD-VISION is dedicated to advancing reliable, robust, and responsible computer vision in the wild: an international forum where researchers and practitioners from academia and industry present and discuss innovative methodologies, benchmarks, datasets, evaluation protocols, and applications for the safe and effective deployment of vision systems. The workshop builds on the success of the Pattern Recognition Special Issue “From Bench to the Wild: Recent Advances in Computer Vision Methods”, which attracted more than 200 submissions, and seeks to consolidate this momentum into a dedicated research community. Beyond presenting cutting-edge research, it will foster critical discussion on benchmark standardization, reproducible evaluation, domain-aware model design, interpretability, efficiency, and responsible deployment. We invite original research contributions on all aspects of the field, including but not limited to the following topics.

PAPER LENGTH: We invite the submission of regular papers; manuscripts must be longer than five pages.

FORMATTING: Manuscripts must comply with the same formatting guidelines required by the main conference, BMVC.

PROCEEDINGS: All accepted papers will be published in the proceedings, together with those of the main conference.

PEER REVIEW: The peer-review process is managed through the Microsoft Conference Management Toolkit (CMT).