SGCN

Dataset and Code for the paper "SVGC-AVA: 360-Degree Video Saliency Prediction with Spherical Vector-Based Graph Convolution and Audio-Visual Attention", TMM

Requirements

• Python3 == 3.8.5
• Pytorch == 1.7.1

Datasets and Pretrained Model

You can download the dataset and models at https://drive.google.com/drive/folders/15WBe_AYPs1geSwft9UUA5x5fQzngXsJF?usp=drive_link

Usages

1.Install the dependencies

pip install -r requirements.txt

2.Generate the dataset

python data_prepare_qin.py
python data_prepare_chao.py

3.Train and test the model

sh main.sh

Citation

If you find this code and dataset is useful for your research, please cite our paper "SVGC-AVA: 360-Degree Video Saliency Prediction with Spherical Vector-Based Graph Convolution and Audio-Visual Attention"