关键词: 关键词: 视频目标分割, 注意力, 融合, 深层语义信息, 浅层位置信息

Abstract: Abstract: In order to solve the problems that the segmentation accuracy and speed are difficult to balance and the algorithm cannot effectively distinguish similar foreground and background in the task of video object segmentation, a semi-supervised video object segmentation algorithm was proposed. Firstly, a pre-generated rough mask was used to process image features, which can achieve more robust features. Secondly, deep semantic information was extracted by attention model. Finally, deep semantic information and shallow position information were fused to obtain accurate segmentation results. The proposed algorithm can achieve higher segmentation accuracy with fast speed and effectively distinguish background and foreground objects. Experiments were conducted on multiple popular datasets. Experimental results demonstrate that the proposed algorithm improves the Jaccard(J) index by 1.8 percentage points and improves the comprehensive evaluation index J&F by 2.3 percentage points compared with Learning Fast and Robust Target Models for Video Object Segmentation (FRTM) on evaluation dataset DAVIS 2016. Meanwhile, on evaluation dataset DAVIS 2017, the proposed algorithm improves the J index by 1.2 percentage points and improves the comprehensive evaluation index J&F by 1.1 percentage points compared with FRTM. The above results fully prove that the proposed algorithm has superior performance in balancing speed and accuracy and effectively distinguishing foreground and background.

Key words: Keywords: video object segmentation, attention, fuse, deep semantic information, shallow position information