Heo Yuk ; Koh Yeong Jun ; Kim Chang-Su

Abstract

We propose a novel guided interactive segmentation (GIS) algorithm for videoobjects to improve the segmentation accuracy and reduce the interaction time.First, we design the reliability-based attention module to analyze thereliability of multiple annotated frames. Second, we develop theintersection-aware propagation module to propagate segmentation results toneighboring frames. Third, we introduce the GIS mechanism for a user to selectunsatisfactory frames quickly with less effort. Experimental resultsdemonstrate that the proposed algorithm provides more accurate segmentationresults at a faster speed than conventional algorithms. Codes are available athttps://github.com/yuk6heo/GIS-RAmap.