NextBestPath

<div align="center"> <h1>NextBestPath: Efficient 3D Mapping of Unseen Environments</h1>

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

<a href="https://arxiv.org/pdf/2502.05378" style="margin-right: 10px;"> <img src="https://img.shields.io/badge/arXiv-Paper-b31b1b?logo=arxiv&logoColor=white" alt="arXiv Paper"> </a> <a href="https://shiyao-li.github.io/nbp/"><img src="https://img.shields.io/badge/Project_Page-green" alt="Project Page"></a>

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

</div>

🌟 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
   

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}
}