Image transformation system and method according to artist style based on artificial neural network

摘要

According to the present invention, a predetermined part set by a user in an image photographed or input by a user, in the style of a work (picture, sculpture, etc.) of a predetermined writer (painter, sculptor, etc.) set by the user, pre-trained artificial neural network It relates to an image transformation system and method according to an artist''s style based on an artificial neural network, performing a style transition by using the image transformation, and outputting it to a user''s terminal. The image modification system according to the style of the present invention includes a camera, which captures a predetermined image, edits the captured image, generates an original image for transforming the style, and selects a style image from images previously stored in the memory unit User terminal to transmit to the server; The original image and the style image received from the user terminal are input to the pre-trained artificial neural network, and the style-modified image output from the artificial neural network is transmitted to the user terminal. Characterized in that it comprises a; a server, which is a transformed image according to a style transition. The original image for transforming the style is composed of the main object image and the background image, and the style image is composed of the main object style image and the background style image. The server inputs the main object image and the main object style image into the artificial neural network, and receives the main object image that is style-modified from the artificial neural network. The image is transformed and transformed according to the style. In addition, the background image and the background style image are input to the artificial neural network, and the style-deformed background image is received from the artificial neural network. It is an image that is transformed and transformed according to the background style image. The style-deformed main object image and the style-deformed background image are synthesized and transmitted as a style-deformed image to the user terminal.