Title: The Worse The Better: Content-Aware Viewpoint Generation Network for Projection-based Point Cloud Quality Assessment
Our work compared the impact of default viewpoint and optimized viewpoint on quality evaluation.
1 data:
We provide viewpoint projection images generated by CAVGN, which can be obtained from this link：BaiduNetdisk
2 dataloder:
Load the corresponding dataset, generate path document and network input: point cloud model+worst quality view label
3 model:
Corresponding model files are stored here.
4 trainer:
Training and testing functions for viewpoint generation networks
5 ViewGenerate:
GetXYZ.py is the file used to generate viewpoints, and we provide a pre-trained model on SJTU-PCQA for extracting tests. Specifically, after generating the viewpoint, a projection image needs to be generated using a projection method.
6 Application:
You can choose various projection based methods and apply them by modifying the original input orthogonal projection image to the projection generated by CAVGN.