search

maturk / dn-splatter

DN-Splatter: Depth and Normal Priors for Gaussian Splatting and Meshing
https://maturk.github.io/dn-splatter/
Apache License 2.0
405 stars 25 forks source link

Getting a warning, normals not found when running ns-train dn-splatter #51

Open rowellz opened 2 months ago

rowellz commented 2 months ago

Hello! Thank you for this awesome project, I am trying to get this running on a monocular estimation dataset. Specifically the sauna one found here. When I get to ns-train dn-splatter [OPTIONS] I get this error WARNING: You have enabled normal supervision with monocular normals but none were found.

For context I am running this in a docker container based off this Dockerfile: 

FROM nerfstudio/nerfstudio:latest

USER root

RUN git clone https://github.com/maturk/dn-splatter

RUN pip install --upgrade pip

RUN cd dn-splatter && \
    pip install -e . 

WORKDIR /workspace

I shell into the container and run these commands in this order:

ns-process-data images --data /workspace/shared/sauna/ --output-dir /workspace/shared/sauna/

python3 /workspace/dn-splatter/dn_splatter/scripts/normals_from_pretrain.py --data-dir /workspace/shared/sauna/ --model-type dsine

Which outputs: Completed generating mono-normals!

I then run python3 /workspace/dn-splatter/dn_splatter/scripts/align_depth.py --data /workspace/shared/sauna/

Which outputs:

Finally I run: ns-train dn-splatter --pipeline.model.use-normal-loss True --pipeline.model.normal-supervision mono coolermap --data /workspace/shared/sauna/ --normals-from pretrained --normal-format opencv

Which outputs:

WARNING: You have enabled normal supervision with monocular normals but none were found.

I suspect the order in which I am running the commands might be incorrect, or perhaps I am missing a command entirely? Regardless, I am a bit surprised by the warning message, the normal maps generated in the /normals_from_pretrain/ dir look good.

Eki87 commented 1 week ago

I got the same problem and make it by adding --load-normals True