Unsupervised segmented video abstraction method based on double-attention mechanism

The embodiment of the invention provides an unsupervised segmented video abstraction method based on a double-attention mechanism, and the method comprises the steps: carrying out the preprocessing of an original video, and segmenting the original video to obtain a video segment group and a shot segment group; respectively inputting the video segment group and the shot segment group into a video abstract model, and processing to obtain weighted video segment features and weighted shot segment features; the weighted video segment features and the weighted shot segment features serve as input, the importance score of each shot is calculated, the shot with the high score, namely the high importance degree, is selected, and a dynamic abstract is generated; constructing a return function, calculating the diversity and representativeness of the dynamic abstract, and training the video abstract model in an unsupervised reinforcement learning mode; according to the method, the importance of the visual content is emph