V-BMS360 CPU

Submission for the ICME Grand Challenge "Salient360! 2018: Visual attention modeling for 360 Images - 2018 edition"

With this archive, two binary can be found:
- salient-image.exe
- salient-video.exe

salient-image is the proposed submission to the image track model type 1: Head motion based saliency model. (This is the BMS360 model, see: https://github.com/Telecommunication-Telemedia-Assessment/GBVS360-BMS360-ProSal)
salient-video contains two models which both address the video track, model type 1: Head motion based saliency model. (This is the V-BMS360 model)

Both model have a manual available through the option --help

This archive contains the CPU version of the V-BMS360 model. This version do not requires a GPU and will only run on CPU. The CPU version of the model use a different optical flow algorithm than the GPU version, therefore some differences can be observed between the model.

It should be noted that in the context of the ICME GC 2018, it is the GPU version of the model which was submitted. Nevertheless, this version should be considered if you do not have a NVidia GPU available as it still show similar performance.

V-BMS360 GPU

Submission for the ICME Grand Challenge "Salient360! 2018: Visual attention modeling for 360 Images - 2018 edition"

With this archive, two binary can be found:
- salient-image.exe
- salient-video.exe

salient-image is the proposed submission to the image track model type 1: Head motion based saliency model. (This is the BMS360 model, see: https://github.com/Telecommunication-Telemedia-Assessment/GBVS360-BMS360-ProSal)
salient-video contains two models which both address the video track, model type 1: Head motion based saliency model. (This is the V-BMS360 model)

Both model have a manual available through the option --help

This archive contains the GPU version of the V-BMS360 model. In this version the NVidia GPU is only used for dense optical flow estimation. Therefore, the model in this version still depend heavily on CPU.