Video Region Annotation with Sparse Bounding Boxes

Video analysis has been moving towards more detailed interpretation (e.g., segmentation) with encouraging progress. These tasks, however, increasingly rely on densely annotated training data both in space and time. Since such annotation is labor-intensive, few video region boundaries exist. This work aims to resolve this dilemma by learning automatically generate for all frames of a from sparsely bounding boxes target regions. We achieve Volumetric Graph Convolutional Network (VGCN), which learns iteratively find keypoints the using spatio-temporal volume surrounding appearance motion. show that global optimization VGCN leads accurate generalizes better. Experimental results three latest datasets (two real one synthetic), including ablation studies, demonstrate effectiveness superiority our method.