search

NVlabs / nvdiffrec

Official code for the CVPR 2022 (oral) paper "Extracting Triangular 3D Models, Materials, and Lighting From Images".
Other
2.09k stars 222 forks source link

DTU Dataset Preparation Script/Steps #69

Open ahmad-bashir9 opened 1 year ago

ahmad-bashir9 commented 1 year ago

Hello, I was working to reproduce results on DTU dataset, but I only have the images. Can you please specify what steps I need to follow to prepare dataset ready for training? It will help a lot. Looking forward to your reply. Thank you.

jmunkberg commented 1 year ago

Hello,

Unfortunately, we did not release the DTU dataset reader and the MLP parameterization of geometry as part of the public release. For the dataset reader for DTU, we adapted code from IDR: https://github.com/lioryariv/idr/blob/main/code/

Some details about the MLP parametrization here: https://github.com/NVlabs/nvdiffrec/issues/9

ahmad-bashir9 commented 1 year ago

Thank you @jmunkberg for your response. Can you please confirm the input for training? I have downloaded your NeRF Synthetic dataset. It has three folders, train, test and val along with the respective .json files. train and val folders have only images while the test have three different files for each image; normal, image and depth. Can you please confirm where these files(in the test folder) came from? Thank you