News
Semantic View Induction for Object-Guided Scene Structuring
Congratulations to lead author Opeyemi Bamigbade and co-authors Mark Scanlon and John Sheppard on their paper Semantic View Induction for Object-Guided Scene Structuring at CVC 2027.
AI-generated summary of the contribution: This work introduces a two-stage framework for uncovering latent scene structure, using Semantic View Induction to discover views from image representations and SceneStruct to assign images to these views while preserving ambiguous samples, with applications in improved retrieval and classification performance.