1 paper
Yongtao Ge, Kangyang Xie, Guangkai Xu +6
Video matting has traditionally been limited by the lack of high-quality ground-truth data. Most existing video matting datasets provide only human-annotated imperfect alpha and fo…