OPTIMIZATION FOR DNN CONPOSITION WITH REAL-TIME INFERENCE IN MOBILE ENVIRONMENT

摘要

And at least one processor implemented to execute computer readable instructions, wherein the at least one processor uses a deep neural network (DNN) -based style transfer model using an image of a specific style to be learned. It includes a learning unit for learning, and the style transfer model provides a system characterized by being a DNN model having a structure in which the number of deep layers is reduced through a transfer learning method using a previously learned result.