oppo-us-research / SpacetimeGaussians

[CVPR 2024] Spacetime Gaussian Feature Splatting for Real-Time Dynamic View Synthesis
https://oppo-us-research.github.io/SpacetimeGaussians-website/
Other
585 stars 43 forks source link

Hello, how do I train my own datasets #46

Open guohaotian0320 opened 5 months ago

guohaotian0320 commented 5 months ago

Hello, how do I train my own datasets?Looking forward to your reply

lizhan17 commented 5 months ago

hi, thank you for your interest in our paper. this repo contains 3 different datasets captured by devices for light-fields. (with three camera rigs and three camera models).

For a custom dataset like (360degree), i think more adaption is needed, especially for the preprocessing and training part. i need some time and gpus for adaption. the situation now is that i am very busy with something else. I can't have time and gpus for custom dataset. Please stay tuned for that.

chensh1127 commented 5 months ago

hi, thank you for your interest in our paper. this repo contains 3 different datasets captured by devices for light-fields. (with three camera rigs and three camera models).

For a custom dataset like (360degree), i think more adaption is needed, especially for the preprocessing and training part. i need some time and gpus for adaption. the situation now is that i am very busy with something else. I can't have time and gpus for custom dataset. Please stay tuned for that.

How do I train my own datasets? ((forward degree)) I tried upload my own dates, but the warning is : FileNotFoundError: [Errno 2] No such file or directory: './data/coser0/poses_bounds.npy' how do I to generate **.npy?

lizhan17 commented 5 months ago

hi, thank you for your interest in our paper. this repo contains 3 different datasets captured by devices for light-fields. (with three camera rigs and three camera models). For a custom dataset like (360degree), i think more adaption is needed, especially for the preprocessing and training part. i need some time and gpus for adaption. the situation now is that i am very busy with something else. I can't have time and gpus for custom dataset. Please stay tuned for that.

How do I train my own datasets? ((forward degree)) I tried upload my own dates, but the warning is : FileNotFoundError: [Errno 2] No such file or directory: './data/coser0/poses_bounds.npy' how do I to generate **.npy?

Hi, custom dataset is not the focus now. Please stay tuned. poses_bounds is not need for costom pipeline.

lizhan17 commented 4 months ago

please check the new command and functions for mulitview cameras without priors.

see https://github.com/oppo-us-research/SpacetimeGaussians/issues/55#issuecomment-2185372592