CCDaS: A Benchmark Dataset for Cartoon Character Detection in Application Scenarios

Deep learning’s achievements in computer vision have poised cartoon character detection (CCD) as a promising tool for intellectual property protection. However, due to the lack of suitable cartoon character datasets, CCD is still a less explored field and there are many issues need to be addressed to meet the demands of practical applications such as merchandise, advertising, and patent examination. In this paper, we introduce CCDaS, a comprehensive benchmark dataset comprising 55,608 images of 524 renowned cartoon characters from 227 works, including cartoons, games, and merchandise. To our knowledge, CCDaS is the most extensive CCD dataset tailored for real-world applications. Alongside, we also provide a CCD algorithm that can achieve accurate detection of animated images in complex practical application scenarios, called multi-path YOLO (MP-YOLO). Experimental results show that our MP-YOLO achieves better detection results on the CCDaS dataset. Comparative and ablation studies further validate the effectiveness of our CCD dataset and algorithm.