Added Dockerfile and requirement

[tauzn-clock]

To help those who wants to play with the repo, I have added a working Dockerfile and requirement.txt for pip install.

Note that the repo needs to be updated before merging.

[HernandoR]

requirements doesn't setup sufficient cuda environment, and may lead to miss understanding for new developers. I would suggest write it in the dockerfile or use a Docker subfolder to place them

[MichaelCurrie]

@tauzn-clock here is a version that uses CUDA 11.7 and puts the requirements inside the Dockerfile to avoid confusion for new developers:

FROM nvidia/cuda:11.7.1-devel-ubuntu20.04
ENV DEBIAN_FRONTEND=noninteractive
ENV DISPLAY=:0
ENV TORCH_CUDA_ARCH_LIST="7.0;7.5;8.0;8.6"
ENV NVIDIA_DRIVER_CAPABILITIES="all"

RUN apt-get update && apt-get install -y \
    lsb-release \
    wget \
    ffmpeg \
    libsm6 \
    libxext6 \
    freeglut3-dev \
    mesa-utils \
    libxmu-dev \
    libxi-dev \
    git \
    python3 \
    python3-dev \
    python3-pip \
    libc6 \
    && apt-get clean \
    && rm -rf /var/lib/apt/lists/*

RUN git clone https://github.com/tauzn-clock/MonoGS/ --recursive
WORKDIR /MonoGS

RUN pip install --upgrade pip
RUN pip install torchaudio --index-url https://download.pytorch.org/whl/cu117 
RUN pip install torchvision --index-url https://download.pytorch.org/whl/cu117
RUN pip install torch --index-url https://download.pytorch.org/whl/cu117

RUN pip install --default-timeout=600 \
    torchmetrics==1.4.1 \
    opencv-python==4.8.1.78 \
    munch==4.0.0 \
    trimesh==4.4.7 \
    evo==1.11.0 \
    open3d==0.18.0 \
    imgviz==1.7.5 \
    PyOpenGL==3.1.7 \
    glfw==2.7.0 \
    PyGLM==2.7.1 \
    wandb==0.17.8 \
    lpips==0.1.4 \
    rich==13.8.0 \
    ruff==0.6.2 \
    plyfile==1.0.3

# Building the submodules requires ninja
RUN pip install ninja --upgrade

RUN pip install submodules/diff-gaussian-rasterization
RUN pip install submodules/simple-knn

[MichaelCurrie]

I have updated this further at #183 so I think this PR 139 should be closed without merging.