|
|
--- |
|
|
extra_gated_prompt: >- |
|
|
You agree to use this dataset exclusively in compliance with the |
|
|
[license](https://building3d.ucalgary.ca/login.php), which is the same as [Building3D dataset](https://building3d.ucalgary.ca/index.php). Pleas don't use mail addresses from free email providers to request data, such as gmail.com, qq.com, web.de, etc. |
|
|
extra_gated_fields: |
|
|
Affiliation: text |
|
|
Name: text |
|
|
Country: country |
|
|
I agree to use this dataset for non-commercial use ONLY: checkbox |
|
|
extra_gated_heading: Acknowledge license to accept the repository |
|
|
extra_gated_description: Our team may take 2-3 days to process your request |
|
|
extra_gated_button_content: Acknowledge license |
|
|
pretty_name: Building3D |
|
|
--- |
|
|
|
|
|
<p style='color: red; font-weight:bold'>1. Pleas don't use mail addresses from free email providers to request data, such as gmail.com, qq.com, web.de, etc.</p> |
|
|
|
|
|
<p style='color: red; font-weight:bold'>2. If you downloaded the dataset before <b>March 5</b>, please download |
|
|
<b>wireframe.zip</b> to update the ground truth of the training set. We have refined the ground truth.</p> |
|
|
|
|
|
# 1. Introduction |
|
|
|
|
|
<p align="justify"><b>Tokyo LoD2 dataset</b> is a subset of the <b>BuildingWorld dataset</b> and |
|
|
primarily represents commercial buildings in <b>Tokyo's downtown area</b>. |
|
|
It consists of the downtown areas of six wards, including <b>Shinjuku, Shibuya, Minato, Itabashi, |
|
|
Nerima, and Chuo</b>. The Tokyo LoD2 dataset provides 33,324 detailed LoD2 building meshes, |
|
|
wireframes, and simulated point clouds. In <b>2nd Building3D Challenge Competition</b>, |
|
|
<b>29,852 buildings</b> are used as the <b>training dataset</b>, while <b>3,472 buildings</b> |
|
|
are used as the <b>testing data</b>. </p> |
|
|
|
|
|
<p align="justify">In the point cloud simulation process, the density of point clouds is set to |
|
|
<b>25 points per square meter </b>. To better approximate real-world sampled point clouds, |
|
|
we introduce <b>random Gaussian perturbations</b> along the <b>normal directions</b> of the point cloud. |
|
|
It is worth noting that we have not introduced random occlusion, which means that the simulated point |
|
|
cloud data is complete and does not have missing corner points or components. |
|
|
Additionally, the point cloud is sampled from textured building models, so it <b>is colored </b> and |
|
|
stored in <b>8-bit format</b>. </p> |
|
|
|
|
|
# 2. Data Analysis |
|
|
|
|
|
## a. Point Cloud analysis |
|
|
|
|
|
|  |  | |
|
|
|:-------------------------------------------------------------------------------------------------------------:|:-----------------------------------------------------------------------------------------------------------------:| |
|
|
| **Box Plot of the number of Point Cloud** | **Point Cloud Distribution After Outlier Removal** | |
|
|
|
|
|
<p align="justify">Approximately 77% of the roof point cloud data contains fewer than 10,000 points, while approximately 60% of data has fewer than 4,096 points. |
|
|
If a fixed number of roof point clouds in training is required, 4,096 maybe a better choice. |
|
|
The number of point clouds in the Building3D dataset can also be set to 4,096, which has been proven to work effectively in the past. |
|
|
However, some buildings with large area have a large amount number of points with a maximum of 227,736 points. |
|
|
Approximately 10% of the data contains more than 20,000 points. These building data may affect the model's performance and, therefore, need to be handled with care. |
|
|
|
|
|
The samples for the test set are drawn uniformly from the entire dataset, which maintains the consistency of the distribution between the training and testing data. |
|
|
</p> |
|
|
|
|
|
## b. Wireframe analysis |
|
|
|
|
|
|  | |
|
|
|:---------------------------------------------------------------------------------------------------------------------:| |
|
|
| **Violin Plots of Wireframe** | |
|
|
|
|
|
|  | |
|
|
|:---------------------------------------------------------------------------------------------------------------------:| |
|
|
| **Histogram of Wireframe** | |
|
|
|
|
|
<p align="justify"> |
|
|
<b>Building3D Dataset:</b> Excluding extremely complex buildings, the maximum number of corner points and edges can reach up to 70. |
|
|
The average corners are about 13, while the average edges are about 15. Interquartile Range (IQR), |
|
|
which is the range between the first quartile (Q1) and the third quartile (Q3), represents the middle 50% of the |
|
|
data distribution. The IQR of corners spans from 6 to 20, while the IQR of edges ranges from 7 to 25. |
|
|
</p> |
|
|
|
|
|
<p align="justify"> |
|
|
<b>Tokyo LoD2 Dataset:</b> Excluding extremely complex buildings, the maximum number of corner points and edges can reach up to 133. |
|
|
The average corners are about 32, while the average edges are about 32. Interquartile Range (IQR), |
|
|
which is the range between the first quartile (Q1) and the third quartile (Q3), represents the middle 50% of the |
|
|
data distribution. The IQR of corners spans from 10 to 58, while the IQR of edges ranges from 11 to 60. |
|
|
</p> |
|
|
|
|
|
|
|
|
# 3. Tokoy LoD2 Data Interactive Visualization |
|
|
|
|
|
|
|
|
<table> |
|
|
<tr> |
|
|
<td align="center"> |
|
|
<img src="https://usm3d.github.io/Building3D/preview_model.png" width="100%"><br> |
|
|
<a href="https://usm3d.github.io/Building3D/building.html"> |
|
|
<img src="https://img.shields.io/badge/View-Building_model-blue?style=for-the-badge"> |
|
|
</a> |
|
|
</td> |
|
|
<td align="center"> |
|
|
<img src="https://usm3d.github.io/Building3D/preview_point_cloud.png" width="100%"><br> |
|
|
<a href="https://usm3d.github.io/Building3D/building_point_cloud.html"> |
|
|
<img src="https://img.shields.io/badge/View-Building_Point_Cloud-blue?style=for-the-badge"> |
|
|
</a> |
|
|
</td> |
|
|
</tr> |
|
|
<tr> |
|
|
<td align="center"> |
|
|
<img src="https://usm3d.github.io/Building3D/preview_roof.png" width="100%"><br> |
|
|
<a href="https://usm3d.github.io/Building3D/roof_point_cloud.html"> |
|
|
<img src="https://img.shields.io/badge/View-Roof_Point_Cloud-blue?style=for-the-badge"> |
|
|
</a> |
|
|
</td> |
|
|
<td align="center"> |
|
|
<img src="https://usm3d.github.io/Building3D/preview_wireframe.png" width="100%"><br> |
|
|
<a href="https://usm3d.github.io/Building3D/wireframe.html"> |
|
|
<img src="https://img.shields.io/badge/View-Wireframe-blue?style=for-the-badge"> |
|
|
</a> |
|
|
</td> |
|
|
</tr> |
|
|
</table> |
|
|
|
