Zero-Shot Visual Question Answering Using Knowledge Graph

Incorporating external knowledge to Visual Question Answering (VQA) has become a vital practical need. Existing methods mostly adopt pipeline approaches with different components for matching and extraction, feature learning, etc. However, such suffer when some component does not perform well, which leads error cascading poor overall performance. Furthermore, the majority of existing ignore answer bias issue—many answers may have never appeared during training (i.e., unseen answers) in real-word application. To bridge these gaps, this paper, we propose Zero-shot VQA algorithm using graph mask-based learning mechanism better incorporating knowledge, present new answer-based splits F-VQA dataset. Experiments show that our method can achieve state-of-the-art performance answers, meanwhile dramatically augment end-to-end models on normal task.