Geodesic Propagation for Semantic Labeling
Qing Li, Xiaowu Chen*, Yafei Song, Yu Zhang, Xin Jin, and Qinping Zhao
This paper presents a semantic labeling framework with geodesic propagation. Under the same framework, three algorithms are proposed, including geodesic propagation (GP), supervised geodesic propagation (SGP) for image and hybrid geodesic propagation (HGP) for video. In these algorithms, we resort to the recognition proposal map and select confident pixels with probability as the initial propagation seeds. From these seeds, the GP algorithm iteratively updates the weights of geodesic distances until the semantic labels are propagated toall pixels. On the contrary, the SGP algorithm further exploits the contextual information to guide the direction of propagation,leading to better performance but higher computational complexity than GP. For video labeling, we further propose the HGP algorithm, in which the geodesic metric is used in both spatial and temporal spaces. Experiments on four public datasets show that our algorithms outperform several state-of-the-art methods.With the geodesic propagation framework, convincing results for both image and video semantic labeling can be obtained.
Qing Li, Xiaowu Chen,
Yafei Song, Yu Zhang, Xin Jin, and Qinping Zhao.Geodesic Propagation for
Xiaowu Chen, Qing Li, Yafei Song, Xin Jin, and Qinping Zhao.Supervised Geodesic Propagation for Semantic Label Transfer,