Boost Video Frame Interpolation via Motion Adaptation

• DOI: 10.48550/arxiv.2306.13933
• Publication Date: 2023-01-01
• Authors: Haoning Wu; Xiaoyun Zhang; Weidi Xie; Ya Zhang; Yanfeng Wang
• Keywords: Motion interpolation; Interpolation; Inter frame

Video frame interpolation (VFI) is a challenging task that aims to generate intermediate frames between two consecutive in video. Existing learning-based VFI methods have achieved great success, but they still suffer from limited generalization ability due the motion distribution of training datasets. In this paper, we propose novel optimization-based method can adapt unseen motions at test time. Our based on cycle-consistency adaptation strategy leverages characteristics among video frames. We also introduce lightweight adapter be inserted into estimation module existing pre-trained models improve efficiency adaptation. Extensive experiments various benchmarks demonstrate our boost performance two-frame models, outperforming state-of-the-art methods, even those use extra input.