Newest papers

VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token

Sangbeom Lim, Seoung Wug Oh, Heeji Yoon, Seungryong Kim, Joon-Young Lee · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26857 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

O computação pesada desses backbones de difusão tem surgido recentemente como uma abordagem líder, aproveitando os preços de difusão de vídeo de pré-treinamento na escala da internet para alcançar uma forte generalização de zero-shot a partir de dados sintéticos limitados. No entanto, o cálculo pesado desses backbones de difusão limita a inferência a um pedaço curto de apenas 6 a 24 quadros, de modo que até mesmo os vídeos comuns abrangem múltiplos pedaços. Porque cada pedaço é matado independentemente da orientação de máscara grosseira, os métodos existentes não conseguem manter a consistência em limites de pedaços, produzindo bordas de flash e transparência descontínua. Nós abordamos isso com um framework de difusão-matagem guiado por máscara constru

Machine translation (Meta M2M100); the English original follows.

Abstract

Diffusion-based video matting has recently emerged as a leading approach, leveraging video diffusion priors from internet-scale pretraining to achieve strong zero-shot generalization from limited synthetic data. However, the heavy compute of these diffusion backbones limits inference to a short chunk of just 6 to 24 frames, so even ordinary videos span multiple chunks. Because each chunk is matted independently from coarse mask guidance, existing methods fail to maintain consistency across chunk boundaries, producing flickering edges and discontinuous transparency. We address this with a mask-guided diffusion-matting framework built on two ideas. The RGB and mask conditioning are represented as separate token streams that interact through full spatio-temporal self-attention, letting the model adaptively combine fine RGB texture with coarse mask layout rather than committing to a fixed channel-wise fusion. On top of this, we propose a novel pair of learnable embeddings, the preserve token and the refine token, that act as a per-frame conditioning interface and enable a sliding-window inference scheme in which already-generated mattes are propagated across chunk boundaries for local temporal consistency, complemented by a globally shared reference anchor at every chunk to prevent error accumulation in long sequences without any additional training. Across standard video matting benchmarks and in-the-wild long footage, our framework outperforms prior diffusion-based methods in both fine-detail accuracy and temporal consistency, and uniquely maintains consistency across chunk boundaries on long videos.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Security token, Computer science, Inference, Generalization, Consistency (knowledge bases), Artificial intelligence

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange