Virtual Multi-Modality Self-Supervised Foreground Matting for Human-Object Interaction | IEEE Conference Publication | IEEE Xplore