MatAnyone

This repository contains the model described in MatAnyone: Stable Video Matting with Consistent Memory Propagation.

Code: https://github.com/pq-yang/MatAnyone

Usage

Requires pip install huggingface_hub.

pip install -q git+https://github.com/pq-yang/MatAnyone

Users will only need to use these two lines to load and instantiate the model.

from matanyone.model.matanyone import MatAnyone
matanyone = MatAnyone.from_pretrained("PeiqingYang/MatAnyone").cuda().eval()

We also provide a stand-alone inference script that could be used as API.

python inference_matanyone_api.py -i <input video> -m <first-frame seg mask>