AI helps make new breakthroughs in art design plagiarism retrieval! Professor Liu Fang's team paper was employed by ACM mm, a multimedia top-level conference

The achievements of Professor Liu Fang's team from the school of design and art of Hunan University 《Understanding and Identifying Artwork Plagiarism with the Wisdom of Designers: A Case Study on Poster Artworks》 By CCF A International Conference ACM MM（30th ACM International Conference on Multimedia） Employment . In recent years , With the vigorous development of artificial intelligence research , The scale and influence of interdisciplinary research are increasing .ACM MM For the field of computer graphics and multimedia of the Chinese computer society A Class meeting , The meeting encourages novelty 、 Innovative theories and methods .

Team member cuishenglan （2020 Master's degree ） And Zhang Mohan （2021 Master's degree ） In Professor Liu Fang （ Corresponding author ） This milestone achievement has been achieved under the guidance of , This is the first time that a master's student from the school of design and art has worked as the first author in CCF A Published papers at international conferences . Relevant research has also been the subject of the national key research and development plan “ Research and development of classic cultural elements and prototype database of Chinese design style ” Project support .

As the first art design works plagiarism retrieval combining artificial intelligence , The team studied the uniqueness and detection difficulties of plagiarism of art and design works , A fusion feature retrieval mechanism based on concept and vision is proposed and implemented , And build a data set for poster plagiarism detection , This achievement will help protect the copyright of design works , At the same time, it supports the repetitive inspection in design teaching , Support the sound development of cultural and creative industries .

The wide spread of digital works and unprecedented interest （ For example, poster illustration works 、NFT）, Making art and design works face more and more risks of plagiarism , This has seriously affected the intellectual property rights of designers and the development of innovation industry . Unlike text and code plagiarism , The design works mainly take the image as the carrier , Plagiarism patterns and characteristics are unknown , As a result, the existing methods can only achieve rough similarity / Same image retrieval effect . Pictured 1 Shown , Although some basic plagiarism cases can be retrieved through simple similar images （A） To identify , But the common problem of design plagiarism is the key object of design 、 style 、 background （B）, There are even complex cases such as collage and theme plagiarism (C).

chart 1 Plagiarism means of different poster design works

Art design works are borrowed or plagiarized , Judging at the legal or professional level is very complex and time-consuming , And there is no available standard to assist in the development of relevant retrieval tools . In response to this challenge , The team combines interdisciplinary features , Through exploratory research with designers , Put forward the evaluation criteria for plagiarism of poster works .

chart 2 Analysis of plagiarism standards for poster design works

It is different from the similar image inspection method based on global visual features , This article combines concepts and visual information to express the multidimensional characteristics of posters . By combining professional design knowledge , This method is suitable for the task of plagiarizing poster retrieval . meanwhile , The team built a large-scale dataset with plagiarism annotations , The proposed model and mechanism achieve a good effect of plagiarism image retrieval .

chart 3 Plagiarize image retrieval mechanism

The innovative achievement is the school of design and art of Hunan University “ New engineering · New design ” One of the new achievements made in the reform and innovation of education and teaching . at present , The plagiarism retrieval tool for art and design works based on this achievement is being developed synchronously , Stay tuned ！

In order to get the result of manuscript review , But it has not been made public