iQIYI iCartoonFace Paper For Cartoon Character Recognition Accepted by Prestigious Multimedia Conference

iQIYI Inc., an innovative market-leading online entertainment service in China, is pleased to announce that its paper on face recognition and detection technology of cartoon characters (the “Paper”), titled Cartoon Face Recognition: A Benchmark Dataset, has been accepted by the Association for Computing Machinery (ACM)’s annual conference on multimedia (ACM Multimedia), a leading academic conference in the field of multimedia technology.

The acceptance of the paper indicates that the value of iQIYI’s iCartoonFace technology has further been acknowledged by the international academic community. Experiments showed that iCartoonFace design framework of the proposed benchmark dataset can effectively improve the efficiency and accuracy of cartoon character recognition.

Recommended AI News: Schneider Electric Solar Expands Its Energy Management Ecosystem with New Smart Edge Devices

ACM Multimedia, the leading international conference in multimedia, focuses on advancing the technological research and applications of images, text, audio, speech, music, sensor and social media.

Rising to the challenges of cartoon character recognition

Although algorithms and applications of real human facial recognition have become increasingly mature, the development of cartoon character recognition technology remains in its infancy. Meanwhile, the scenarios requiring cartoon character recognition technology are growing rapidly as the animation industry booms.

Aside from video-oriented structured analysis, there is also robust demand for cartoon character recognition technology in image search, ad recognition and other use cases. However, as cartoon characters are more complex than human characters, they pose much greater technical challenges. Unlike human characters, who are fairly consistent in appearance, cartoon faces are far more varied in terms of color and texture as well as their potential for small inter-class differences and large intra-class variations.

A high-quality cartoon dataset is key to realizing technology breakthroughs and relevant applications of cartoon character recognition. However, most existing cartoon recognition datasets are riddled with issues such as data insufficiency and sound contamination. In the Paper, iQIYI’s team presents a new multi-character training framework for cartoons and real people. The Paper also introduces a semi-automatic dataset construction framework and the world’s largest cartoon character detection and recognition dataset with manual annotation — iCartoonFace. iCartoonFace contains over 400,000 high-quality real-scene images of over 5,000 characters. It can be applied to structured analysis, intelligent video-editing, image searching, ad recognition and even identification of satirical cartoons characters.

Recommended AI News: NetSPI Adds to Leadership Team to Support Continued Focus on Customer Success

Currently, cartoon recognition technology enabled by iCartoonFace is being used in several of the Company’s innovative offerings and software development work. For example, iQIYI’s AIWorks feature supports the automatic collection of cartoon character elements and generates AI-enabled video mashups. The AI Radar ‘Qiguan’ feature allows users to identify cartoon characters when viewing animated content.

iQIYI drives innovation in pursuit of delivering optimal user experiences

iQIYI is dedicated to developing cutting-edge technologies and exploring the applications of new technologies in entertainment. In the last two years, the Company successfully held two multimodal video character recognition challenges which helped to significantly advance the accuracy of multimodal character recognition.

Most recently, IQYI partnered with the International Joint Conference on Artificial Intelligence and the Pacific Rim International Conference on Artificial Intelligence to hold the first large-scale cartoon character recognition competition in China. This competition attracted a wide range of talent, with 500 teams from top universities such as Peking University, Fudan University, Zhejiang University, Shanghai Jiao Tong University, and University of Chinese Academy of Sciences, as well as well-known technology companies such as Alibaba, Tencent, Baidu, JD.com and SenseTime. Looking ahead, iQIYI will continue to strengthen its research and innovation as it strives to constantly explore the next generation of tech-driven applications and experiences.

Recommended AI News: Frost & Sullivan Examines Seven VR and AR Application Areas Boosted by Global 5G Deployment