Video Analysis System Using Deep Learning Algorithms

摘要

Detection of video duplicates is an active field of research, motivated by the protection of intellectual property, the fight against piracy or the tracing of the origin of reused video segments. In this work, a method for the detection of duplicate videos is proposed and implemented, making use of deep learning methods and techniques typical of the field of information recovery. This method has been evaluated with a data set usually used in the field, with which high average accuracies, above 85%, have been obtained. The effect of the different layers of the convolutional neural network used by the algorithm, the aggregation mechanisms that can be used on them, and the influence of the recovery model have been studied, finding a set of parameters that optimize the overall accuracy of the system.