Counting challenging crowds robustly using a multi-column multi-task convolutional neural network

作者：

Highlights：

•

• A novel density map is proposed to encode location and detailed information.

• A multi-column CNN is improved through minimizing per-scale loss.

• Multi-tasks learning is used to improve the counting performance.

摘要

•A novel density map is proposed to encode location and detailed information.•A multi-column CNN is improved through minimizing per-scale loss.•Multi-tasks learning is used to improve the counting performance.

论文关键词：Crowd counting,Multi-column CNN,Multi-task,Per-scale loss,Density map

论文评审过程：Received 25 October 2017, Revised 7 March 2018, Accepted 7 March 2018, Available online 12 March 2018, Version of Record 26 March 2018.

论文官网地址：https://doi.org/10.1016/j.image.2018.03.004