KR-20260067677-A - System and method for building dataset for copyright protection in game broadcasting

KR20260067677AKR 20260067677 AKR20260067677 AKR 20260067677AKR-20260067677-A

Abstract

A dataset construction system for copyright protection in game broadcasting according to the present invention is characterized by comprising: a generation unit that generates an image and metadata similar to an actual game scene and determines whether the generated image is a ‘True’ image; and a storage unit that stores the image and metadata when the image generated by the generation unit is determined to be ‘True’.

Inventors

장성일
홍두표
김현수
조용준
신동명
김민수

Assignees

엘에스웨어(주)

Dates

Publication Date: 20260513
Application Date: 20241106

Claims (8)

A generation unit that generates an image and metadata similar to an actual game scene and determines whether the generated image is a 'True'image; and A dataset construction system for copyright protection in game broadcasting, characterized by including a storage unit that stores the image and metadata when the image generated by the above-mentioned generation unit is determined to be 'True'.
In claim 1, The above generating unit is, A dataset construction system for copyright protection in game broadcasting, composed of the structure of a Generative Adversarial Network, which is a generative artificial intelligence, and characterized by the learning of a generative model (generator) and a discriminative model (discriminator).
In claim 2, A dataset construction system for copyright protection in game broadcasting, characterized in that the generator of the above-mentioned generation unit receives a randomly generated noise vector as input and generates an image and metadata similar to a game scene, and the discriminator receives the image generated from the generator and the actual game scene image as input and determines whether the image is 'True' or 'Fake'.
In claim 1, The above storage unit is, A dataset construction system for copyright protection in game broadcasting, characterized by storing the image and metadata when the image generated by the above-mentioned generation unit is determined to be 'True', while separating the image file and metadata to store and manage them separately for efficient data processing and management.
A step of generating an image and metadata similar to an actual game scene and determining whether the generated image is a 'True'image; and A method for constructing a dataset for copyright protection in game broadcasting, characterized by including a step of saving the image and metadata when the generated image is determined to be 'True'.
In claim 5, A method for constructing a dataset for copyright protection in game broadcasting, characterized by being composed of the structure of a Generative Adversarial Network, which is a generative artificial intelligence, and having a generative model (generator) and a discriminative model (discriminator) learn.
In claim 6, A method for constructing a dataset for copyright protection in game broadcasting, characterized in that a generator receives a randomly generated noise vector as input and generates an image and metadata similar to a game scene, and a discriminator receives the image generated from the generator and an actual game scene image as input and determines whether the image is 'True' or 'Fake'.
In claim 5, A method for constructing a dataset for copyright protection in game broadcasting, characterized by storing the image and metadata when the generated image is determined to be 'True', and separating the image file and metadata to store and manage them separately for efficient data processing and management.

Description

System and method for building dataset for copyright protection in game broadcasting The present invention relates to a technology for constructing a dataset for copyright protection in game broadcasting. Game streaming broadcasts, which are being utilized as a new marketing tool for games, require that a licensing agreement be concluded and managed in advance between the game developer (original creator) and the streamer. If revenue is generated through broadcasting without a contract, this may lead to copyright infringement of the game work. However, for numerous games (original works), there are difficulties in concluding and managing licensing agreements between game developers (original authors) and streamers. Currently, game developers rely on a method of self-monitoring and manually sanctioning broadcasts that violate their regulations. From a technical perspective, to determine or verify contract compliance during game broadcasting, a model that analyzes game scenes to identify relevant game information can be utilized, and such a model requires a sufficient amount of training data. FIG. 1 is a configuration block diagram illustrating a dataset construction system for copyright protection in game broadcasting according to the present invention. Figure 2 is a reference diagram illustrating a method of storing an image generated in a generation unit in a storage unit. Figure 3 is a flowchart illustrating a method for constructing a dataset for copyright protection in game broadcasting according to the present invention. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. The embodiments of the present invention are provided to more fully explain the invention to those skilled in the art, and the following embodiments may be modified in various different forms, and the scope of the invention is not limited to the following embodiments. Rather, these embodiments are provided to make the disclosure more faithful and complete and to fully convey the spirit of the invention to those skilled in the art. The terms used herein are for describing specific embodiments and are not intended to limit the invention. As used herein, the singular form may include the plural form unless the context clearly indicates otherwise. As used herein, the term "and/or" includes any one of the listed items and all combinations of one or more thereof. Generally, if the amount of training data for a game information identification model is small, there is a high possibility that the model will overfit to the low diversity of the training data during training and fail to meet the performance level required in real-world environments. Therefore, to improve the performance of a game information identification model, it is necessary to secure a sufficient amount of data that can increase the scale and diversity of the game scene training data. In order for game developers and streamers to resolve the difficulties of managing existing license agreements, it is necessary to prioritize improving the performance level of models for identifying game information before developing technology to verify compliance with the agreement. The learning dataset construction system proposed in this invention generates images and metadata similar to actual game scenes and determines whether the generated images are ‘True’ images. To this end, it is configured with the structure of a Generative Adversarial Network, a generative artificial intelligence, and a generative model (generator) and a discriminative model (discriminator) are trained. Additionally, if a generated image is determined to be ‘True’, the corresponding image and metadata are stored. FIG. 1 is a configuration block diagram illustrating a dataset construction system for copyright protection in game broadcasting according to the present invention. Referring to FIG. 1, a dataset construction system for copyright protection in game broadcasting according to the present invention includes a generation unit (100) and a storage unit (200). The generation unit (100) generates an image and metadata similar to an actual game scene and determines whether the generated image is a ‘True’ image. The generation unit (100) is composed of the structure of a Generative Adversarial Network, which is a generative artificial intelligence, and the generation model (generator) and the discrimination model (discriminator) learn. The generator of the generation unit (100) receives a randomly generated noise vector as input and generates an image and metadata similar to a game scene, and the discriminator receives the image generated from the generator and the actual game scene image as input and determines whether the image is ‘True’ or ‘Fake’. Game scene images refer to scenes representing specific situations within the game, such as combat scenes, story scenes, and the main screen. Image metadata is generated based on parameters, c