Unsupervised video abstraction method and system based on hybrid twin mask automatic encoder

The invention discloses an unsupervised video abstraction method and an unsupervised video abstraction system based on a mixed twin mask automatic encoder, which directly quantify the importance of each shot in a video through a mode of recovering shots after masking, improve the accuracy of shot importance evaluation, and improve the video abstraction accuracy. Compared with a fitting result of an artificial abstract, the method has the advantages that the problem that a conventional unsupervised video abstract method based on a generative adversarial model is unstable in training is effectively solved, the generated abstract result is more stable, and the model training of the method does not depend on complex artificial annotation, so that the method is more convenient to implement. Compared with a supervised method, the method has higher feasibility and can be effectively applied to the fields of video classification, retrieval and the like. 本发明公开了一种基于混合孪生掩码自动编码器的无监督视频摘要方法及系统，本方法通过掩码后恢复镜头的方式，直接量化视频中每个镜头的重