Resilience through Scene Context in Visual Referring Expression Generation

arXiv:2404.12289v1 Announce Type: new
Abstract: Scene context is well known to facilitate humans' perception of visible objects. In this paper, we investigate the role of context in Referring Expression Generation (REG) for objects in images, where existing research has often focused on distractor contexts that exert pressure on the generator. We take a new perspective on scene context in REG and hypothesize that contextual information can be conceived of as a resource that makes REG models more resilient and facilitates …

abstract arxiv context cs.cl generator humans images objects paper perception research resilience role through type visual