Splatter Image运行笔记

文章标题：Splatter Image: Ultra-Fast Single-View 3D Reconstruction

1. 环境配置

下载Splatter Image代码

git clone https://github.com/szymanowiczs/splatter-image.git

创建环境

conda create --name splatter-image python=3.8

激活环境

conda activate splatter-image

安装torch、torchvision、torchaudio

pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu124

安装其它依赖

pip install -r requirements.txt

pip install rembg onnxruntime

下载3D Gaussian Splatting代码 

git clone https://github.com/graphdeco-inria/gaussian-splatting --recursive

安装diff-gaussian-rasterization

cd gaussian-splatting

pip install submodules/diff-gaussian-rasterization

2. 下载数据

import os
from huggingface_hub import snapshot_download


# 使用cache_dir参数，将模型/数据集保存到指定“本地路径”
snapshot_download(repo_id="szymanowiczs/splatter-image-v1", repo_type="dataset",
                  cache_dir="/home/xxxxx/Downloads",
                  local_dir_use_symlinks=False, resume_download=True,
                  token='hf_***')

3. 下载预训练模型

import os
from huggingface_hub import snapshot_download


# 使用cache_dir参数，将模型/数据集保存到指定“本地路径”
snapshot_download(repo_id="szymanowiczs/splatter-image-v1", repo_type="model",
                  cache_dir="/home/xxxxx/Downloads",
                  local_dir_use_symlinks=False, resume_download=True,
                  token='hf_***')

4. 训练

5. 推理