Abstract
As more and more digitized manga (Japanese comics) books are available, efficient and effective access to manga is urgently needed. Among various elements of manga, character's face plays one of the most important roles in access and retrieval. We propose a deep neural network method to do manga face detection, which is a challenging but relatively unexplored topic. Given a manga page, we first find candidate regions based on the selective search scheme. Three convolutional neural networks are then proposed to detect manga faces of various appearance. We extract information from the entire object region and several local regions, and integrate multi-scale information in an early fusion manner or a late fusion manner. The proposed methods are evaluated based on a large-scale benchmark. Convincing performance compared to the state-of-the-art face detection modules designed for human faces is demonstrated.
Original language | English |
---|---|
Pages (from-to) | 62-72 |
Number of pages | 11 |
Journal | Pattern Recognition |
Volume | 86 |
DOIs | |
Publication status | Published - 2019 Feb |
All Science Journal Classification (ASJC) codes
- Software
- Signal Processing
- Computer Vision and Pattern Recognition
- Artificial Intelligence