Flexible Interactive Guided Image Matting

Proceedings Paper Computer Science, Artificial Intelligence

MatteFormer: Transformer-Based Image Matting via Prior-Tokens

GyuTae Park et al.

Summary: The proposed MatteFormer model utilizes the trimap information in transformer blocks, using prior tokens for self-attention mechanism, and achieves state-of-the-art performance on commonly used image matting datasets.

2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR) (2022)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Robust High-Resolution Video Matting with Temporal Guidance

Shanchuan Lin et al.

Summary: The method introduced is an efficient, real-time, high-resolution human video matting approach that outperforms previous methods. By utilizing a recurrent architecture and a novel training strategy, it enhances temporal coherence and matting quality.

2022 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2022) (2022)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Deep Video Matting via Spatio-Temporal Alignment and Aggregation

Yanan Sun et al.

Summary: This paper proposes a deep learning-based video matting framework with a novel spatio-temporal feature aggregation module (ST-FAM) to effectively align and aggregate information across spatial scales and temporal frames. Additionally, a lightweight interactive trimap propagation network is introduced to eliminate frame-by-frame trimap annotations. The framework outperforms conventional video matting and deep image matting methods in the presence of multi-frame temporal information.

2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021 (2021)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence