arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.08834cs.CVcs.LG
收藏

GSVNet:用于视频快速语义分割的引导空间变化卷积

GSVNet: Guided Spatially-Varying Convolution for Fast Semantic Segmentation on Video

National Yang Ming Chiao Tung University(台湾阳明交通大学)

Shih-Po Lee, Si-Cun Chen, Wen-Hsiao Peng

摘要:

本文研究了视频的快速语义分割,视频分割通常要求实时甚至比实时更快的处理。用于保存由特征提取引起的计算的一个常见配方是传播少数选定关键帧的特征。然而,快速图像分割的最新进展使得这些解决方案不那么有吸引力。为了利用快速图像分割进一步的视频分割,我们提出了一个简单而有效的传播框架。具体来说,我们在1/8缩小的图像空间中进行轻量级流估计,以在分割空间中进行时间扭曲。此外,我们引入了一个引导的空间变化的卷积融合分割来自以前和当前帧,以减轻传播误差,并使轻量级的特征提取非关键帧。在Cityscapes和CamVid上的实验结果表明,该算法在视频分割上达到了最佳的精度和吞吐量的平衡。

英文摘要:

This paper addresses fast semantic segmentation on video.Video segmentation often calls for real-time, or even fasterthan real-time, processing. One common recipe for conserving computation arising from feature extraction is to propagate features of few selected keyframes. However, recent advances in fast image segmentation make these solutions less attractive. To leverage fast image segmentation for furthering video segmentation, we propose a simple yet efficient propagation framework. Specifically, we perform lightweight flow estimation in 1/8-downscaled image space for temporal warping in segmentation outpace space. Moreover, we introduce a guided spatially-varying convolution for fusing segmentations derived from the previous and current frames, to mitigate propagation error and enable lightweight feature extraction on non-keyframes. Experimental results on Cityscapes and CamVid show that our scheme achieves the state-of-the-art accuracy-throughput trade-off on video segmentation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑