Giter VIP home page Giter VIP logo

leonwu0108 / neusurf Goto Github PK

View Code? Open in Web Editor NEW
38.0 4.0 0.0 6.13 MB

[AAAI'24] NeuSurf: On-Surface Priors for Neural Surface Reconstruction from Sparse Input Views

Home Page: https://alvin528.github.io/NeuSurf/

License: MIT License

Python 92.92% Cuda 6.10% C++ 0.98%
3d-reconstruction computer-graphics computer-vision neural-rendering multi-view-reconstruction nerf surface-reconstruction

neusurf's Introduction

NeuSurf

Implementation of AAAI'24 paper NeuSurf: On-Surface Priors for Neural Surface Reconstruction from Sparse Input Views

Overview

Installation

Our code is implemented in Python 3.10, PyTorch 2.0.0 and CUDA 11.7.

  • Install Python dependencies
conda create -n neusurf python=3.10
conda activate neusurf
pip install torch==2.0.0 torchvision==0.15.1
pip install -r requirements.txt
  • Compile C++ extensions
cd extensions/chamfer_dist
python setup.py install

Dataset

Data structure:

data
|-- DTU_pixelnerf
    |-- <case_name, e.g. dtu_scan24>
        |-- cameras_sphere.npz
        |-- pcd
            |-- <case_name>.ply
        |-- cam4feat
            |-- pair.txt
            |-- cam_00000000_flow3.txt
            |-- cam_00000001_flow3.txt
            ...
        |-- image
            |-- 000000.png
            |-- 000001.png
            ...
        |-- mask
            |-- 000.png
            |-- 001.png
            ...
|-- DTU_sparseneus
|-- blendedmvs_sparse

You can directly download the processed data here.

Running

  • Training
CUDA_VISIBLE_DEVICES=0
python exp_runner.py --mode train --conf ./confs/dtu.conf --case <case_name, e.g. dtu_scan24>
  • Extract mesh
CUDA_VISIBLE_DEVICES=0
python exp_runner.py --mode validate_mesh --conf ./confs/dtu.conf --case <case_name> --is_continue

Citation

If you find our work useful in your research, please consider citing:

@inproceedings{huang2024neusurf,
  title={NeuSurf: On-Surface Priors for Neural Surface Reconstruction from Sparse Input Views},
  author={Huang, Han and Wu, Yulun and Zhou, Junsheng and Gao, Ge and Gu, Ming and Liu, Yu-Shen},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  number={3},
  pages={2312--2320},
  year={2024}
}

Acknowledgement

This implementation is based on CAP-UDF, D-NeuS and Vis-MVSNet. Thanks for these great works.

neusurf's People

Contributors

alvin528 avatar leonwu0108 avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar

neusurf's Issues

Release Mesh Results

Can you please release the meshes for your method and other methods you've compared?

Data Preprocess

May I ask how I can run through my own data? Is there a formalized data preprocessing scheme? Thank you for your great work.

About Preparing my own dataset

Hi! An interesting work! thanks for your great work.
When I follow MVSDF(Vis-MVSNet) to prepare my own dataset, I meet a question :
when I want to run test.py(in Vis-MVSNet)
python test.py --data_root ./dtu_scan75 --dataset_name general --num_src 4 --max_d 256 --resize 1280,720 --crop 1280,720 --load_path pretrained_model/vis --write_result --result_dir ./output
TypeError: forward() missing 3 required positional arguments: 'sample', 'depth_nums', and 'interval_scales'
my data structure as:

  • <dtu_scan75>
    • - 00000000_cam.txt - ......
    • - 00000000.jpg - ......
    • <images_col>
      • 0.png
      • ......
    • <sparse_col>
      • cameras.txt (which is generated by COLMAP)
      • image.txt (which is generated by COLMAP)
      • points3D.txt (which is generated by COLMAP)
    • pair.txt
      can you give me some advice?Thanks.

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.