Could not recognize scene type! about gaussianeditor HOT 4 CLOSED

According to the readme file a base docker image should be like this

FROM  nvidia/cuda:11.8.0-devel-ubuntu22.04

RUN apt update && apt install -y python3 python3-pip  git ninja-build libeigen3-dev libgl1-mesa-glx  libglib2.0-0
RUN pip install --upgrade pip
RUN pip install torch==2.0.1+cu118 torchvision==0.15.2+cu118 --extra-index-url https://download.pytorch.org/whl/cu118
COPY requirements.txt /app

WORKDIR /app
ENV TCNN_CUDA_ARCHITECTURES=86
RUN pip install -r requirements.txt
RUN pip install mediapipe


WORKDIR /app/extern
RUN git clone https://github.com/heheyas/viser  && pip install -e viser
WORKDIR /app

sub modules should be installed in running container because of error that no cuda runtime during docker build.

pip install -e gaussiansplatting/submodules/diff-gaussian-rasterization
pip install -e gaussiansplatting/submodules/simple-knn

from gaussianeditor.

Please check your colmap_dir
This is caused by sparse is not in colmap_dir

from gaussianeditor.

Thanks, I checked the download.sh again (I downloaded files manually before),
unzip the 2 model zip files into one folder , it works.

Now struggling with the CPU only issue. It seems like a torch or driver version problem

Please check your colmap_dir This is caused by sparse is not in colmap_dir

GaussianEditor/gaussiansplatting/scene/camera_scene.py

Line 26 in f00e40d

if os.path.exists(os.path.join(source_path, "sparse")):

from gaussianeditor.

It seems the same "CPU mode only" problem in docker:
IDEA-Research/GroundingDINO#8 (comment)

from gaussianeditor.