IROS-2025-Challenge-Nav

<div id="top" align="center">
<img src="https://github.com/InternRobotics/InternNav/raw/main/challenge/demo.gif" width=60% >
</div>

# IROS-2025-Challenge-Nav Dataset

## Dataset Summary 📖

This dataset includes the R2R dataset and the InteriorNav dataset, constructed from Matterport3D scanned environments and InteriorNav(kujiale) high-quality modeled environments, respectively, with corresponding navigation trajectories and language instructions.

### Trajectory Statistics by Subset

| Dataset | Train | Val Seen | Val Unseen | Test Unseen |
|------------------|------------------------|-------------------|---------------------|------------------------|
| VLN-PE-R2R | 8,679 (stair-filtered) | 778 | 1,839 | 3,408 |
| InteriorNav | 649 | 44 | 99 | 165 |
| **Total** | **9,328** | **822** | **1,938** | **3,573** |

# Get started 🔥
## Download the Dataset
```
# Make sure git-lfs is installed (https://git-lfs.com)
git lfs install

git clone https://huggingface.co/datasets/InternRobotics/IROS-2025-Challenge-Nav

# If you want to clone without large files - just their pointers
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/InternRobotics/IROS-2025-Challenge-Nav
```

## Dataset Structure 📁

```
vln_pe
├── raw_data/ # JSON files defining tasks, navigation goals, and dataset splits
│ └── r2r/
│ ├── mini/
│ │ └── mini.json.gz # For quick Model and Environments validation
│ ├── train/
│ ├── val_seen/
│ │ └── val_seen.json.gz
│ ├── val_unseen/
│ │ └── val_unseen.json.gz
│ └── embeddings.json.gz
└── traj_data # training sample data for two types of scenes
├── interiornav/
│ ├── kujiale_xxxx.tar.gz
│ └── ...
└── r2r/
├── traj_index/
│ ├── data/
│ ├── meta/
│ └── videos/
└── ...
```

# License and Citation
All the data and code within this repo are under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/). Please consider citing our project if it helps your research.

```BibTeX
@misc{contributors2025internroboticsrepo,
title={IROS-2025-Challenge-Nav Colosseum},
author={IROS-2025-Challenge-Nav Colosseum contributors},
howpublished={\url{https://github.com/InternRobotics/InternNav/tree/main/challenge}},
year={2025}
}
```

<div id="top" align="center">
<img src="https://github.com/InternRobotics/InternNav/raw/main/challenge/demo.gif" width=60% >
</div>

# IROS 2025挑战赛导航数据集（IROS-2025-Challenge-Nav Dataset）

## 数据集概述 📖

本数据集包含R2R数据集与InteriorNav数据集（InteriorNav Dataset），分别构建自Matterport3D扫描环境以及库嘉乐（kujiale）高精度建模环境，配套对应导航轨迹与自然语言指令。

### 各子集轨迹统计

| 数据集名称 | 训练集 | 可见验证集 | 不可见验证集 | 不可见测试集 |
|------------------|------------------------|-------------------|---------------------|------------------------|
| VLN-PE-R2R | 8,679（已过滤楼梯场景） | 778 | 1,839 | 3,408 |
| InteriorNav | 649 | 44 | 99 | 165 |
| **总计** | **9,328** | **822** | **1,938** | **3,573** |

## 快速上手 🔥
## 数据集下载

# 请确保已安装Git LFS（https://git-lfs.com）
git lfs install

git clone https://huggingface.co/datasets/InternRobotics/IROS-2025-Challenge-Nav

# 若仅需克隆大文件指针而非完整文件
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/InternRobotics/IROS-2025-Challenge-Nav

## 数据集结构 📁

vln_pe
├── raw_data/ # 定义任务、导航目标与数据集划分的JSON文件
│ └── r2r/
│ ├── mini/
│ │ └── mini.json.gz # 用于快速验证模型与环境的mini.json.gz
│ ├── train/
│ ├── val_seen/
│ │ └── val_seen.json.gz
│ ├── val_unseen/
│ │ └── val_unseen.json.gz
│ └── embeddings.json.gz
└── traj_data # 两类场景的训练样本数据
├── interiornav/
│ ├── kujiale_xxxx.tar.gz
│ └── ...
└── r2r/
├── traj_index/
│ ├── data/
│ ├── meta/
│ └── videos/
└── ...

# 许可与引用
本仓库内所有数据与代码均采用[CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/)许可协议。若本数据集对您的研究有所帮助，请考虑引用我们的成果。

BibTeX
@misc{contributors2025internroboticsrepo,
title={IROS-2025-Challenge-Nav Colosseum},
author={IROS-2025-Challenge-Nav Colosseum contributors},
howpublished={url{https://github.com/InternRobotics/InternNav/tree/main/challenge}},
year={2025}
}

搜集汇总

背景与挑战

背景概述

IROS-2025-Challenge-Nav数据集整合了R2R和InteriorNav两个子集，专为视觉语言导航（VLN）研究设计，包含丰富的导航轨迹和语言指令数据。数据集基于Matterport3D和InteriorNav环境构建，适用于机器人导航算法的训练与验证。