Skip to content

Latest commit

 

History

History
33 lines (29 loc) · 1.19 KB

README.md

File metadata and controls

33 lines (29 loc) · 1.19 KB
Raw

Sparse Depth Prior Generation

We use Colmap to generate sparse depth. Installation instructions can be found here. Run the following files to generate sparse depth priors for the respective datasets for all the three input configurations.

cd src/prior_generators/sparse_depth/
python DepthEstimator01_RealEstate.py
python DepthEstimator02_NeRF_LLFF.py
cd ../../../

Running the above files creates a new directory data/databases/<DATABASE_NAME>/data/all/estimated_depths, which contains three sub-directories named DE02,DE03,DE04 for LLFF dataset and DE12,DE13,DE14 for RealEstate dataset corresponding to two, three and four input-view settings. Each of these directories will contain multiple sub-directories, one for every scene in the dataset. The following tree shows an exmaple.

data/databases/NeRF_LLFF/data/all/estimated_depths
|--DE02
|  |--fern
|  |  |--estimated_depths_down4
|  |  |  |--0006.csv
|  |  |  |--0013.csv
|  |  |--EstimatedBounds.csv
|  |--flower
|  ...  
|--DE03
|  |--fern
|  ...
|--DE04
|  |--fern
|  ...

Acknowledgements

Parts of the code are borrowed from DS-NeRF codebase.