Scanpath and saliency prediction on 360 degree images

作者：

Highlights：

• Introduction of saliency volumes to capture the temporal nature of scan-paths.

• The SaltiNet architecture to generate scan-paths from a deep neural network.

• This work was the best scanpath solution at ICME 2017 (De Abreu et al., 2017).

• A similar architecture suitable for saliency map prediction.

摘要

•Introduction of saliency volumes to capture the temporal nature of scan-paths.•The SaltiNet architecture to generate scan-paths from a deep neural network.•This work was the best scanpath solution at ICME 2017 (De Abreu et al., 2017).•A similar architecture suitable for saliency map prediction.

论文关键词：Deep learning,Machine learning,Saliency,Scanpath,Visual attention

论文评审过程：Received 21 September 2017, Revised 13 June 2018, Accepted 13 June 2018, Available online 23 June 2018, Version of Record 30 October 2018.

论文官网地址：https://doi.org/10.1016/j.image.2018.06.006