The love of animation and cartoons is a global phenomenon that crosses age, gender and ethnic divides. Even if you’re not a cosplay enthusiast who wigs up for anime conventions, you are most likely a fan of more than a few anime shorts and films. According to SimilarWeb traffic analysis, US anime streaming site Crunchyroll attracts around 80 million visits per month, ranking it #15 among global streaming websites with an impressive 60 percent the traffic of category leader YouTube.
In a bid to combine anime and cartoon culture with machine learning, a research team from China’s leading video streaming service iQIYI told Synced it is introducing a novel large unconstrained cartoon dataset, “iCartoonFace.”
Baidu-owned iQiyi went public last year in the US and is one of China’s largest platforms for watching animation and cartoons. In step with China’s rising obsession with anime culture, the company is investing heavily in licensing foreign anime TV shows and developing domestic anime productions.
IQiyi positions itself as a tech-driven platform with artificial Intelligence implemented throughout its business process, in content creation, production, distribution, customer experience and so on. In the new iCartoonFace paper, company researchers believe their dataset will have applications in tasks such as recognition research, cartoon person modeling and image classification.
The ICartoonFace dataset contains 68,312 manually annotated images with 2,639 identities from 739 anime and cartoon albums. The samples were extracted from publicly available images on the Internet and videos hosted on the iQiyi site.
It’s expected the iCartoonFace dataset will be released soon. Its associated paper iCartoonFace: A Benchmark of Cartoon Person Recognition is currently under a review process. Synced is covering the story and will continue to update readers with the latest news.
Author: Yuqing Li | Editor: Michael Sarazen