Congested scene classification via efficient unsupervised feature learning and density estimation

作者：

Highlights：

• Assemble a new data set to serve as the platform for crowd analysis.

• Present an efficient feature learning and selection to generate robust features.

• Propose a novel density feature pooling to encode the density clue.

摘要

Highlights•Assemble a new data set to serve as the platform for crowd analysis.•Present an efficient feature learning and selection to generate robust features.•Propose a novel density feature pooling to encode the density clue.

论文关键词：Computer vision,Unsupervised feature learning,Scene classification,Density estimation,Spherical k-means,Feature pooling

论文评审过程：Received 4 August 2015, Revised 8 March 2016, Accepted 15 March 2016, Available online 24 March 2016, Version of Record 12 April 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.03.020