Sovereign 8s.is mirror of standardgalactic/NextBestPath
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-04-14 13:34:14 -04:00
.idea Initial commit 2024-12-05 11:22:21 -05:00
configs Initial commit 2025-06-07 22:36:21 -04:00
macarons update 2025-11-05 08:06:07 -05:00
next_best_path update 2025-11-18 10:00:22 -05:00
vis_blender add vis script 2026-04-14 13:34:14 -04:00
.gitignore Initial commit 2024-12-05 11:22:21 -05:00
environment.yml Initial commit 2025-06-07 22:36:21 -04:00
README.md add vis script 2026-04-14 13:34:14 -04:00
test_nbp_planning.py Initial commit 2025-06-07 22:36:21 -04:00
train_nbp.py Initial commit 2025-06-07 22:36:21 -04:00

NextBestPath: Efficient 3D Mapping of Unseen Environments

Shiyao Li, Antoine Guédon, Clémentin Boittiaux, Shizhe Chen, Vincent Lepetit

arXiv Paper Project Page

A method for generating the next-best-path for efficient active mapping, along with a new benchmark tailored for complex indoor environments.

🌟 If you find our work helpful, please consider giving a ⭐️ to this repository and citing our paper!

🗺️ Project Overview

NextBestPath (NBP) is a novel method for next-best-path planning in 3D scene exploration. Unlike previous methods, NBP is designed to directly maximize mapping efficiency and coverage along the camera trajectory.

This repository contains:

  • A simulator based on PyTorch3D and Trimesh
  • Functions for generating ground truth point clouds from meshes and evaluating reconstructed point clouds
  • Scripts for testing and training NBP models on AiMDoom dataset.
@inproceedings{li2025nextbestpath,
  title={NextBestPath: Efficient 3D Mapping of Unseen Environments},
  author={Shiyao Li and Antoine Guedon and Cl{\'e}mentin Boittiaux and Shizhe Chen and Vincent Lepetit},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025},
  url={https://openreview.net/forum?id=7WaRh4gCXp}
}

Updates

  • [June, 2025] Release the training and test code
  • Todo: Release the models of MACARONS and the corresponding scripts

Quick Start

Prerequisites

First, ensure you have conda installed, then set up the environment:

# Clone this repository
git clone https://github.com/shiyao-li/NextBestPath.git
cd NextBestPath

# Create and activate conda environment
conda env create -f environment.yml
conda activate exploration

Installation

  1. Download the AiMDoom Dataset

    Download the complete dataset from Google Drive:

    • AiMDoom dataset (4 difficulty levels)
    • The toolkit and code to build AiMDoom dataset: Github_link
  2. Download and set up model weights

    Download NBP models from Google Drive, and put them under the ./weights/nbp/ folder.

    Place the downloaded NBP model weights in the following structure:

    ./weights/nbp/
    ├── AiMDoom_simple_best_val.pth  
    ├── AiMDoom_normal_best_val.pth  
    ├── AiMDoom_hard_best_val.pth  
    └── AiMDoom_insane_best_val.pth
    

Usage

  1. Configs

    All config files are under the ./configs/ folder.

  2. Test NBP method

    python test_nbp_planning.py
    
  3. Train NBP models

    python train_nbp.py
    

Visualization

For visualization-related code, please refer to vis_blender/.

Citation

If you use this work in your research, please cite:

@inproceedings{li2025nextbestpath,
  title={NextBestPath: Efficient 3D Mapping of Unseen Environments},
  author={Shiyao Li and Antoine Guedon and Cl{\'e}mentin Boittiaux and Shizhe Chen and Vincent Lepetit},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025},
  url={https://openreview.net/forum?id=7WaRh4gCXp}
}