Decoupled global-local collaborative network for visual question answering
Visual Question Answering (VQA) aims to achieve cross-modal semantic understanding through joint modeling of visual content and natural language. Although existing attention-based approaches effectively align features, they struggle to simultaneously accommodate global semantic modeling and local fine-grained perceptio...