Overview / Usage

A dataset of 360-degree video saliency maps. The dataset can be used to further investigate human behavior or facilitate 360-degree streaming

Methodology / Approach

Human head tracking is collected from public dataset. The head tracking was collected by tracking users head orientation during 360-degree free viewing session. Users in these experiment user Head Mounted Device.

The collected head tracking was aggregated and project onto equirectangular frame to create saliency maps. The saliency maps show indicate the spatial region with level of interested

Technologies Used

Python numpy, sklearn
Head Mounted Devices: Occulus Rift, Dreamview
Deep learning framework: tensorflow, keras, caffe

Repository

https://github.com/phananh1010/PanoSalNet