StylizedFacePoint: Facial Landmark Detection for Stylized Characters

Shengran Cheng, Chuhang Ma, Ye Pan

Abstract: Facial landmark detection forms the foundation for numerous face-related tasks. Recently, this field has gained substantial attention and made significant advancements. Nonetheless, detecting facial landmarks for stylized characters still remains a challenge. Existing approaches, which are mostly trained on real-human face datasets, struggle to perform well due to the structural variations between real and stylized characters. Additionally, a comprehensive dataset for analyzing stylized characters' facial features is lacking. This study proposes a novel dataset, the Facial Landmark Dataset for Stylized Characters (FLSC), which contains 2674 images and 4086 faces. These data is selected from 16 cartoon video clips, together with 98 landmarks per image, labeled by professionals. Besides, we propose StylizedFacePoint: a deep-learning-based method for stylized facial landmark detection that outperforms the existing approaches. This method has also proven to work well for characters with styles outside the training domain. Moreover, we outline two primary types of applications for our dataset and method. For each, we provide a detailed illustrative example.

Primary Subject Area: [Content] Media Interpretation

Relevance To Conference:

This work's multimedia/multimodal objects include images, videos, and stylized characters. Its contributions to the realm of multimedia and multimodal processing can be summarized as follows:
1. Establishment of a novel dataset, the Facial Landmark Dataset for Stylized Characters (FLSC): Recognizing facial landmarks in stylized characters poses a challenge due to their geometric and textural diversity. The creation of FLSC aims to bridge this gap by providing a comprehensive dataset tailored to address this specific need.
2. Breakthrough in facial landmark detection for stylized characters: Leveraging deep learning techniques, we have developed a landmark detection method for stylized characters. Our approach surpasses existing detectors and methodologies, empowering researchers to analyze and manipulate stylized character faces with greater precision and efficiency.
3. Advancement in landmark-related applications of stylized characters: We outline possible applications facilitated by our landmark detector and offer two detailed examples. By following this framework, our work has the potential to elevate the quality and realism of stylized multimedia content creation.

Submission Number: 2072