Efficient Semantic-Guidance High-Resolution Video Matting

Yue Yu*, Ding Li, Yulin Yang

*此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

Video matting has made significant progress in trimap-based field. However, researchers are increasingly interested in auxiliary-free matting because it is more useful in real-world applications. We propose a new efficient semantic-guidance high-resolution video matting network for human body. We apply the convolutional network as the backbone while also employing the transformer in the encoder, which is used to utilize semantic features, while ensuring that the network is not overly bloated. In addition, a channel-wise attention mechanism is introduced in the decoder to improve the representation of semantic feature. In comparison to the current state-of-the-art methods, the method proposed in this paper achieves better results while maintaining the speed and efficiency of prediction. We can complete the real-time auxiliary-free matting for high-resolution video (4K or HD).

源语言英语
主期刊名Advances in Computer Graphics - 40th Computer Graphics International Conference, CGI 2023, Proceedings
编辑Bin Sheng, Lei Bi, Jinman Kim, Nadia Magnenat-Thalmann, Daniel Thalmann
出版商Springer Science and Business Media Deutschland GmbH
143-154
页数12
ISBN(印刷版)9783031500688
DOI
出版状态已出版 - 2024
活动40th Computer Graphics International Conference, CGI 2023 - Shanghai, 中国
期限: 28 8月 20231 9月 2023

出版系列

姓名Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
14495
ISSN(印刷版)0302-9743
ISSN(电子版)1611-3349

会议

会议40th Computer Graphics International Conference, CGI 2023
国家/地区中国
Shanghai
时期28/08/231/09/23

指纹

探究 'Efficient Semantic-Guidance High-Resolution Video Matting' 的科研主题。它们共同构成独一无二的指纹。

引用此