Research and implementation of visual question and answer system based on deep learning

Abstract With the development and improvement of deep learning technology, its application practice in modal data (image, speech, text) has been achieved tremendously. In this paper, based on neural network modality class model learning, we analyze adaptation to visual question answer system, propose a gated attention mechanism, construct question, prediction mechanism adapted recurrent transfer model. To address problem low accuracy complex problems, inference module is built using so that can extract features improve capability By predicting answers through semantic information about text elements images, correlations across modalities, inference, advances natural language processing computational vision have led improved learning-based quiz models. Multiple sets experiments show models with capabilities questions significantly higher than other existing methods.