Electronic device for generating user-preferred content, and operating method therefor
Provided are an electronic device for generating user-preferred content, and an operating method of the electronic device. The operating method may include: obtaining first content including at least one of an image or audio; detecting at least one feature to be used to change a style of content, by applying the first content to a feature detection model; generating second content including an image-audio pair by applying the first content and the at least one feature of the first content to a neural style transfer model, wherein the neural style transfer model is trained to determine an image or audio to be included in the second content, and generate the second content by changing a style of the first content; and outputting the second content.
1 . An operating method of an electronic device, the operating method comprising:
obtaining first content including at least one of an image or audio;
detecting at least one feature to be used to change a style of the first content, by applying the first content to a feature detection model;
applying the first content and the at least one feature to a neural style transfer (NST) model, wherein the NST model is trained to generate a second content based on the first content and the at least one feature applied to the NST model by changing the style of the first content, and the second content includes an image audio pair in which an image or audio is determined by the NST model; and
outputting the second content,
wherein
the NST model includes a first sub-network configured to change a style of an image and a second sub-network configured to change a style of audio, and
at least some layers of the first sub-network are connected to at least some layers of the second sub-network, for weight sharing between the first sub-network and the second sub-network.
2 . The operating method of claim 1 , further comprising:
obtaining a usage history of the electronic device,
wherein the applying further includes applying the usage history of the electronic device to the NST model, and the NST model is further trained to generate the second content based on the usage history of the electronic device applied to the NST model.
3 . The operating method of claim 2 , wherein the usage history of the electronic device includes at least one of a history related to applications that have been executed by the electronic device, a history related to content that has been reproduced by the electronic device, and a history related to external sources that have been connected to and used by the electronic device.
4 . The operating method of claim 1 , further comprising:
obtaining content viewing environment information including at least one of illuminance and chromaticity of a space where the electronic device is located,
wherein the applying further includes applying the content viewing environment information to the NST model, and the NST model is further trained to generate the second content based on the content viewing environment information applied to the NST model.
5 . The operating method of claim 1 , wherein the detecting of the at least one feature includes, based on the first content including an image, detecting one or more main objects from the image.
6 . The operating method of claim 5 , wherein the detecting of the at least one feature includes determining the one or more main objects based on a user input.
7 . An electronic device comprising:
a memory storing one or more instructions; and
a processor configured to execute the one or more instructions stored in the memory to:
obtain first content including at least one of an image or audio,
detect at least one feature to be used to change a style of the first content, by applying the first content to a feature detection model,
apply the first content and the at least one feature to a neural style transfer (NST) model, wherein the NST model is trained to generate a second content based on the first content and the at least one feature applied to the NST model by changing the style of the first content, and the second content includes an image audio pair in which an image or audio is determined by the NST model, and
output the second content,
wherein the NST model includes:
a first sub-network configured to change a style of an image and a second sub-network configured to change a style of audio, and
at least some layers of the first sub-network are connected to at least some layers of the second sub-network, for weight sharing between the first sub-network and the second sub-network.
8 . The electronic device of claim 7 , wherein
the processor is further configured to execute the one or more instructions to:
obtain a usage history of the electronic device, and
apply the usage history of the electronic device to the NST model, and
the NST model is further trained to generate the second content based on the usage history of the electronic device applied to the NST model.
9 . The electronic device of claim 8 , wherein the usage history of the electronic device includes at least one of a history related to applications that have been executed by the electronic device, a history related to content that has been reproduced by the electronic device, and a history related to external sources that have been connected to and used by the electronic device.
10 . The electronic device of claim 7 , wherein
the processor is further configured to execute the one or more instructions to:
obtain content viewing environment information including at least one of illuminance and chromaticity of a space where the electronic device is located, and
apply the content viewing environment information to the NST model, and
the NST model is further trained to generate the second content based on the content viewing environment information applied to the NST model.
11 . The electronic device of claim 7 , wherein the processor is further configured to execute the one or more instructions to:
based on the first content including an image, detect one or more main objects from the image.
12 . The electronic device of claim 11 , wherein the processor is further configured to execute the one or more instructions to:
determine the one or more main objects based on a user input.
13 . A non-transitory computer-readable recording medium having recorded thereon a program for executing, on a computer, the operating method of claim 1 .