Deep learning methods are increasingly used in cross-media retrieval research because of their powerful performance. However, due to the large number of parameters and the large amount of calculation of the neural network model, the speed of cross-media retrieval is limited accordingly. In view of the above problems, this paper applies the compressed convolutional neural network VGG-16 to cross-media retrieval, and obtains a better retrieval speed. The specific method is to extract image features using channel-pruned deep convolutional neural n...