Test3R: Learning to Reconstruct 3D at Test Time

Test3R: Learning to Reconstruct 3D at Test Time
Yuheng Yuan, Qiuhong Shen, Shizun Wang, Xingyi Yang, Xinchao Wang
xML Lab, National University of Singapore

Getting Started

Installation

1. Clone Test3R.

git clone --recursive https://github.com/nopQAQ/Test3R.git
cd Test3R

2. Create the environment, here we show an example using conda.

conda create -n test3r python=3.11 cmake=3.14.0
conda activate test3r 
conda install pytorch torchvision pytorch-cuda=12.1 -c pytorch -c nvidia  # use the correct version of cuda for your system
pip install -r requirements.txt

3. Optional, compile the cuda kernels for RoPE (as in CroCo v2).

cd croco/models/curope/
python setup.py build_ext --inplace
cd ../../../

Evaluation

Datasets

Please follow Spann3R and Robustmvd to prepare 7scenes, Neural-RGBD, DTU and ETH3D datasets.

Multi-view Reconstruction

bash eval/mv_recon/run.sh

ToDo List

• More functions and demos.

• Evaluation code on Robustmvd.

• Implementation on VGGT.

Related Work

Our code, data preprocessing pipeline, and evaluation scripts are based on several awesome repositories. Welcome to also check out these awesome works, including but not limited to:

• 3D Geometric Foundation Models: DUSt3R / CUT3R / MASt3R / Spann3R / VGGT

• Visual Prompt: VPT / SPT

BibTeX

@misc{yuan2025test3rlearningreconstruct3d,
      title={Test3R: Learning to Reconstruct 3D at Test Time}, 
      author={Yuheng Yuan and Qiuhong Shen and Shizun Wang and Xingyi Yang and Xinchao Wang},
      year={2025},
      eprint={2506.13750},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2506.13750}, 
}