PMNI: Pose-free Multi-view Normal Integration for Reflective and Textureless Surface Reconstruction

Project page: https://pmz-enterprise.github.io/PMNI/

Correction

Equation (13) in the main text of the paper is incorrect. It should be:

$$\alpha \equiv \frac{\mathbf{z}^{r} \cdot \mathbf{z}^{ni}}{\mathbf{z}^{ni} \cdot \mathbf{z}^{ni}}.$$

Quick Start

Code was tested on Ubuntu 20.04 using Python 3.8, PyTorch 2.1.0, and CUDA 11.8 on an Nvidia RTX4090 (24GB). Before started, please ensure CUDA is installed in your environment. It is required by tiny-cuda-nn.

You should see something like the following after typing `nvcc --version`

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:33:58_PDT_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

Clone the repository and prepare the conda environment:

git clone https://github.com/pmz-enterprise/PMNI
cd PMNI 
. ./create_env.sh
pip install "git+https://github.com/facebookresearch/pytorch3d.git@stable" # install pytorch3d
pip install pyexr # install pyexr

Data

Our data is available at :

Google drive: https://drive.google.com/drive/folders/1lIw3eXFPiL-h9nY5k6LpYKQHvakFE_fq?usp=drive_link

Train

CUDA_VISIBLE_DEVICES=X python exp_runner.py --conf config/xx.conf --obj_name xx

For your own data

We strongly recommend using DUSt3R to provide a reasonable initial pose.

Video

Our supplementary video is available at: https://youtu.be/hLIZG24m1Wo

Bibtex

@inproceedings{pmni,
title = {PMNI: Pose-free Multi-view Normal Integration for Reflective and Textureless Surface Reconstruction},
author = {Mingzhi, Pei and Xu, Cao and Xiangyi, Wang and Heng, Guo and Zhanyu, Ma },
year = {2025},
booktitle = CVPR,
}

Acknowledgement

Our implementation is built from SuperNormal, NoPe-NeRF, BARF.