A deep generative directed network for scene depth ordering.

Journal of Visual Communication and Image Representation(2019)

引用 1|浏览71
暂无评分
摘要
•We propose deep generative directed-network (DGDN) to unify two complementary edgel and regional cues conflicts.•DGDN plugs the HMF-like structure onto the top of deep dense network.•HMF model use bi-direction inference to update the state of depth ordering.•We further design an EM-like procedure to learn the optimal parameters.
更多
查看译文
关键词
Deep generative directed-network,Depth ordering,Hidden Markov field,DenseNet
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要