VideoMaMa: Mask-Guided Video Matting via Generative Prior
This work addresses the problem of scalable video matting for real-world applications by leveraging generative priors and pseudo-labeling, though it is incremental in building on existing segmentation and diffusion models.
The paper tackles the challenge of generalizing video matting models to real-world videos by introducing VideoMaMa, which uses pretrained video diffusion models to convert coarse masks into accurate alpha mattes, achieving strong zero-shot generalization with synthetic training. It also creates the MA-V dataset with over 50K annotated videos and fine-tunes SAM2 to outperform existing models on in-the-wild videos.
Generalizing video matting models to real-world videos remains a significant challenge due to the scarcity of labeled data. To address this, we present Video Mask-to-Matte Model (VideoMaMa) that converts coarse segmentation masks into pixel accurate alpha mattes, by leveraging pretrained video diffusion models. VideoMaMa demonstrates strong zero-shot generalization to real-world footage, even though it is trained solely on synthetic data. Building on this capability, we develop a scalable pseudo-labeling pipeline for large-scale video matting and construct the Matting Anything in Video (MA-V) dataset, which offers high-quality matting annotations for more than 50K real-world videos spanning diverse scenes and motions. To validate the effectiveness of this dataset, we fine-tune the SAM2 model on MA-V to obtain SAM2-Matte, which outperforms the same model trained on existing matting datasets in terms of robustness on in-the-wild videos. These findings emphasize the importance of large-scale pseudo-labeled video matting and showcase how generative priors and accessible segmentation cues can drive scalable progress in video matting research.