Description
Implement a current Gaussian Splatting Method in our Photogrammetry Webservice Pix2Model
Last time we checked, Gaussian Splatting variants were not usable for us. They were either very bad or took much too long. In the meantime, hundreds of new variants have been developed. There are a few that should get a chance to show their quality.
Tasks
- Check current Gaussian Splatting Methods' applicability for real-world scenes
- Implement the best one in our Photogrammetry Webservice
A hot candidate for Gaussian Splatting Reconstruction: https://anttwo.github.io/matcha/
Our Photogrammetry Webservice: https://pix2model.cg.tuwien.ac.at/
Requirements
- Knowledge of English language (source code comments and final report should be in English)
- Knowledge of C++
- Basic knowledge of machine learning
Bonus:
- Knowledge of CUDA
- Knowledge of Dockerunderstanding
Environment
The project should be implemented in our Photogrammetry Webservice Pix2Model. This is a Flask Server on Ubuntu with processing steps running in Docker containers.
Responsible
For more information please contact Philipp Erler.