ZIM: Zero-Shot Image Matting for Anything
Published in IEEE/CVF International Conference on Computer Vision (ICCV 2025, Highlight), 2025
Recommended citation: Kim et al. "ZIM: Zero-Shot Image Matting for Anything." Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 23828–23838. https://arxiv.org/abs/2411.00626
Authors
Beomyoung Kim, Chanyong Shin, Joonhyun Jeong, Hyungsik Jung, Se-Yun Lee, Sewhan Chun, Dong-Hyun Hwang, and Joonsang Yu
Overview
ZIM extends zero-shot segmentation toward detailed image matting. The work introduces a process for converting segmentation annotations into matte labels, together with a hierarchical pixel decoder and prompt-aware attention. Evaluation uses MicroMat-3K and considers applications including inpainting and 3D reconstruction.
ICCV 2025 · Highlight