Video Classification Technology Based on Deep Learning
Mengyun Liu · 2020
It takes a large amount of time for artificial video classification and analysis, and the processing efficiency is low, so it is difficult to better meet the practical needs for the era of big data. Therefore, it is extremely urgent to form an intelligent video analysis and classification method, which is worth exploring. In this paper, experiments on feature extraction, feature fusion and similarity measurement are carried out to determine methods to improve video classification accuracy. The results show that the classification accuracy decreases with the increase of pyramid layers. The performance of the cascade SRU method is better, which can improve the classification accuracy. In addition, two-level cascading coding can better complete the extraction of effective information in depth characteristics and promote the accuracy of classification. The classification accuracy of the network increases, after the characteristic aggregation is accomplished by means of average pooling.