search

facebookresearch / nonrigid_nerf

Open source repository for the code accompanying the paper 'Non-Rigid Neural Radiance Fields Reconstruction and Novel View Synthesis of a Deforming Scene from Monocular Video'.
Other
361 stars 40 forks source link

Missing precomputed.json output #13

Open weders opened 2 years ago

weders commented 2 years ago

Hi @edgar-tr,

Thanks a lot for providing the code for this work! I'd like to run it on sequential data that has been calibrated using the provided preprocessing tool.

Unfortunately, the precomputed.json file that is required for poses and everything is not generated. Do you know what could be the problem? I somehow cannot find the code that produces that file in the preprocess.py.

Thank your very much!

edgar-tr commented 2 years ago

The pipeline with the precomputed.json is only for the example sequence. Running the preprocessing code on other sequences does not produce such a json file. Instead, preprocessing.py should produce poses_bounds.npy etc. similar to https://github.com/yenchenlin/nerf-pytorch

The main processing in preprocess.py happens in the single line here: https://github.com/facebookresearch/nonrigid_nerf/blob/main/preprocess.py#L336 This calls gen_poses. For scenes other than the dummy scene, it's easiest to use the LLFF preprocessing code as described in the fourth point here: https://github.com/facebookresearch/nonrigid_nerf/tree/5187abfe9b9e30e45cb87087f94ff1e00fb4450c#installation

I hope that helps!