Method for solving video question and answer task based on multi-mode progressive attention model

Abstract

The embodiment of the invention provides a method for solving a video question and answer task based on a multi-mode progressive attention model. The method comprises the following steps of: 1, respectively extracting various modal features for various modal information in a video question and answer task; 2, performing preliminary attention on the extracted modal features by using the problem, calculating corresponding weight scores, and performing iterative attention on the important modal features by using the problem so as to position the modal features most relevant to the problem; 3, realizing cross-modal fusion of the features by using a multi-modal fusion algorithm, paying attention to multi-modal fusion representation of the video by using the problem, and finding out important video features related to the problem; and 4, fusing part of effective output results of the model for generating answers. Compared with an existing video question-answering solution, the video frame or the video picture area related to the question can be positioned more accurately. Compared with a traditional method, the effect obtained in the video question and answer task is better.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of video question answering, and more specifically, embodiments of the present invention relate to a method for solving video question answering tasks based on a multimodal progressive attention model. Background technique [0002] In recent years, video question answering is a challenging nascent field that has attracted much attention from researchers. This task requires the model to understand the semantic information between the video and the question, and generate an answer based on this semantic information. Since open-ended questions require the model to automatically generate natural language answers, open-ended questions are more difficult types of questions in video question answering tasks at this stage. [0003] In question answering tasks, video information is more complex than image information. Video is an image sequence with strong temporal dynamics, and there are a la...

AI Technical Summary

Problems solved by technology

[0005] In this context, the embodiments of the present invention expect to provide a method for solving video question answering tasks based on a multimodal progressive attention model, so as to overcome the problem that the prior art cannot provide more accurate answers for video question answering tasks

Images