What is the best way for extracting meaningful attributes from pictures?
作者：
Highlights：
• We proposed a metric to automatically measure visual attribute meaningfulness.
• The metric is based on subspace interpolation on a decision boundary manifold.
• Meaningfulness distance is computed via approximated distance on the manifold.
• An improved metric calibration method is developed based on the in-depth analysis.
• We present extensive experiments and analysis on four popular attribute datasets.
摘要
Highlights•We proposed a metric to automatically measure visual attribute meaningfulness.•The metric is based on subspace interpolation on a decision boundary manifold.•Meaningfulness distance is computed via approximated distance on the manifold.•An improved metric calibration method is developed based on the in-depth analysis.•We present extensive experiments and analysis on four popular attribute datasets.
论文关键词：Visual attribute,Meaningfulness metric,Attribute discovering,Semantic content
论文评审过程：Received 8 April 2016, Revised 17 October 2016, Accepted 31 October 2016, Available online 4 November 2016, Version of Record 3 December 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.10.034
