SpatialLM-Testset

# SpatialLM Testset

[Project page](https://manycore-research.github.io/SpatialLM) | [Paper](https://arxiv.org/abs/2506.07491) | [Code](https://github.com/manycore-research/SpatialLM)

We provide a test set of 107 preprocessed point clouds and their corresponding GT layouts, point clouds are reconstructed from RGB videos using [MASt3R-SLAM](https://github.com/rmurai0610/MASt3R-SLAM). SpatialLM-Testset is quite challenging compared to prior clean RGBD scan datasets due to the noises and occlusions in the point clouds reconstructed from monocular RGB videos.

<table style="table-layout: fixed;">
<tr>
<td style="text-align: center; vertical-align: middle; width: 25%"> <img src="./figures/a.jpg" alt="exmaple a" width="100%" style="display: block;"></td>
<td style="text-align: center; vertical-align: middle; width: 25%"> <img src="./figures/b.jpg" alt="exmaple b" width="100%" style="display: block;"></td>
<td style="text-align: center; vertical-align: middle; width: 25%"> <img src="./figures/c.jpg" alt="exmaple c" width="100%" style="display: block;"></td>
<td style="text-align: center; vertical-align: middle; width: 25%"> <img src="./figures/d.jpg" alt="exmaple d" width="100%" style="display: block;"></td>
</tr>
</tr>
</table>

## Folder Structure

Outlines of the dataset files:

```bash
project-root/
├── pcd/*.ply # Reconstructed point cloud PLY files
├── layout/*.txt # GT FloorPlan Layout
├── benchmark_categories.tsv # Category mappings for evaluation
└── test.csv # Metadata CSV file with columns id, pcd, layout
```

## Usage

Use the [SpatialLM code base](https://github.com/manycore-research/SpatialLM/tree/main) for reading the point cloud and layout data.

```python
from spatiallm import Layout
from spatiallm.pcd import load_o3d_pcd

# Load Point Cloud
point_cloud = load_o3d_pcd(args.point_cloud)

# Load Layout
with open(args.layout, "r") as f:
layout_content = f.read()
layout = Layout(layout_content)
```

## Visualization

Use `rerun` to visualize the point cloud and the GT structured 3D layout output:

```bash
python visualize.py --point_cloud pcd/scene0000_00.ply --layout layout/scene0000_00.txt --save scene0000_00.rrd
rerun scene0000_00.rrd
```

# SpatialLM测试集

[项目页面](https://manycore-research.github.io/SpatialLM) | [论文](https://arxiv.org/abs/2506.07491) | [代码](https://github.com/manycore-research/SpatialLM)

本数据集提供107组预处理后的点云及其对应的真值（Ground Truth，简称GT）布局，所有点云均通过[MASt3R-SLAM](https://github.com/rmurai0610/MASt3R-SLAM)从RGB视频中重建得到。相较于此前的干净RGB-D扫描数据集，SpatialLM测试集因单目RGB视频重建的点云存在噪声与遮挡问题，难度显著更高。

<table style="table-layout: fixed;">
<tr>
<td style="text-align: center; vertical-align: middle; width: 25%"><img src="./figures/a.jpg" alt="示例a" width="100%" style="display: block;"></td>
<td style="text-align: center; vertical-align: middle; width: 25%"><img src="./figures/b.jpg" alt="示例b" width="100%" style="display: block;"></td>
<td style="text-align: center; vertical-align: middle; width: 25%"><img src="./figures/c.jpg" alt="示例c" width="100%" style="display: block;"></td>
<td style="text-align: center; vertical-align: middle; width: 25%"><img src="./figures/d.jpg" alt="示例d" width="100%" style="display: block;"></td>
</tr>
</tr>
</table>

## 数据集文件夹结构

数据集文件结构如下：
bash
project-root/
├── pcd/*.ply # 重建点云PLY文件
├── layout/*.txt # 真值平面布局文件
├── benchmark_categories.tsv # 评估用类别映射表
└── test.csv # 元数据CSV文件，包含id、pcd、layout三列

## 使用方法

可使用[SpatialLM代码库](https://github.com/manycore-research/SpatialLM/tree/main)读取点云与布局数据。
python
from spatiallm import Layout
from spatiallm.pcd import load_o3d_pcd

# 加载点云
point_cloud = load_o3d_pcd(args.point_cloud)

# 加载布局
with open(args.layout, "r") as f:
layout_content = f.read()
layout = Layout(layout_content)

## 可视化方法

可使用`rerun`工具对点云与真值结构化3D布局输出结果进行可视化：
bash
python visualize.py --point_cloud pcd/scene0000_00.ply --layout layout/scene0000_00.txt --save scene0000_00.rrd
rerun scene0000_00.rrd