Semi-supervised learning for refining image annotation based on random walk model
作者：
Highlights：
• A unified framework for refining annotation by fusing GMM with random walk.
• A semi-supervised learning is used to enhance the quality of training data.
• Gaussian mixture model is learned to estimate initial annotations.
• Integrating label and visual similarities of images associated with the labels.
• Random walk is implemented on graph to further mine the semantic correlation.
摘要
•A unified framework for refining annotation by fusing GMM with random walk.•A semi-supervised learning is used to enhance the quality of training data.•Gaussian mixture model is learned to estimate initial annotations.•Integrating label and visual similarities of images associated with the labels.•Random walk is implemented on graph to further mine the semantic correlation.
论文关键词：Automatic image annotation,Semi-supervised learning,Gaussian mixture model,Expectation-maximization,Random walk,Image retrieval
论文评审过程：Received 12 October 2013, Revised 8 August 2014, Accepted 29 August 2014, Available online 9 September 2014.
论文官网地址：https://doi.org/10.1016/j.knosys.2014.08.023
