autonomousvision / giraffe

This repository contains the code for the CVPR 2021 paper "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields"
https://m-niemeyer.github.io/project-pages/giraffe/index.html
MIT License
1.23k stars 160 forks source link

About the scale and transform, rotation #6

Closed zhangqianhui closed 3 years ago

zhangqianhui commented 3 years ago

Congrats for the best paper.

I want to know about how to attain the these parameters for new dataset.

for example:

scale_range_min: [0.21, 0.21, 0.21]
scale_range_max: [0.21, 0.21, 0.21]
translation_range_min: [0., 0., 0.]
translation_range_max: [0., 0., 0.]
rotation_range: [0.375, 0.625]

Thanks!

m-niemeyer commented 3 years ago

Hi @zhangqianhui , thank you for your interest in our project!

You are right that you need to adjust these ranges to your own data. How you need to adjust them highly depends on the dataset you plan to use - and the closer you match what's correct, the better the results will be. To get a feeling for how to adjust things, I would suggest to have a look at the different config files (e.g. cars, celebA, church, etc.).

Best of luck for your research!