Datasets:
Tasks:
Image Segmentation
Modalities:
Image
Formats:
imagefolder
Size:
1K - 10K
Tags:
building
License:
update the dataset card
Browse files
README.md
CHANGED
|
@@ -6,4 +6,162 @@ tags:
|
|
| 6 |
- building
|
| 7 |
size_categories:
|
| 8 |
- 1K<n<10K
|
| 9 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
- building
|
| 7 |
size_categories:
|
| 8 |
- 1K<n<10K
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
# Building Contour Detection and Height Estimation Problem
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
## Dataset Summary
|
| 19 |
+
|
| 20 |
+
The **building_height_estimation** dataset is a collection of satellite images with annotated building footprints (polygons) and corresponding building heights.
|
| 21 |
+
It is designed for the **joint tasks of building contour detection (segmentation)** and **height estimation (regression)** from monocular aerial images.
|
| 22 |
+
|
| 23 |
+
* **Source & Owner**: The dataset originates from the [AlgoTester Building Contour Detection & Height Estimation Contest](https://algotester.com/en/ContestProblem/DisplayWithFile/135254).
|
| 24 |
+
* **Hugging Face Host**: [MElHuseyni / building_height_estimation](https://huggingface.co/datasets/MElHuseyni/building_height_estimation).
|
| 25 |
+
* **License**: MIT
|
| 26 |
+
* **Tags**: `building`, `segmentation`, `regression`, `remote sensing`
|
| 27 |
+
|
| 28 |
+

|
| 29 |
+
|
| 30 |
+
*Example of a dataset sample: a 512Γ512 satellite image with annotated building footprints (polygon overlays) and their corresponding height values.*
|
| 31 |
+
|
| 32 |
+
---
|
| 33 |
+
|
| 34 |
+
## Dataset Description
|
| 35 |
+
|
| 36 |
+
Each sample consists of:
|
| 37 |
+
|
| 38 |
+
* **Image**: RGB satellite image (512Γ512 px).
|
| 39 |
+
* **Annotations**: A LabelMe-like JSON containing:
|
| 40 |
+
|
| 41 |
+
* `points`: Polygon vertices (x, y) marking building footprints
|
| 42 |
+
* `group_id`: Numeric building height (meters)
|
| 43 |
+
|
| 44 |
+
The dataset supports training and evaluation for:
|
| 45 |
+
|
| 46 |
+
* Building footprint extraction
|
| 47 |
+
* Height estimation from monocular imagery
|
| 48 |
+
* Joint segmentation + regression architectures
|
| 49 |
+
|
| 50 |
+
---
|
| 51 |
+
|
| 52 |
+
## Dataset Structure
|
| 53 |
+
|
| 54 |
+
```
|
| 55 |
+
/
|
| 56 |
+
βββ images/ # Training images
|
| 57 |
+
β βββ img0001.png
|
| 58 |
+
β βββ ...
|
| 59 |
+
β
|
| 60 |
+
βββ ground_truth_files/ # Training labels (JSON)
|
| 61 |
+
β βββ img0001.json
|
| 62 |
+
β βββ ...
|
| 63 |
+
β
|
| 64 |
+
βββ test_images/ # Test set (no ground truth provided)
|
| 65 |
+
βββ test0001.png
|
| 66 |
+
βββ ...
|
| 67 |
+
```
|
| 68 |
+
|
| 69 |
+
**Example Label (JSON):**
|
| 70 |
+
|
| 71 |
+
```json
|
| 72 |
+
{
|
| 73 |
+
"shapes": [
|
| 74 |
+
{
|
| 75 |
+
"points": [[316, 486], [307, 510], [312, 512]],
|
| 76 |
+
"group_id": 9
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"points": [[416, 457], [435, 446], [421, 423], [402, 434]],
|
| 80 |
+
"group_id": 7
|
| 81 |
+
}
|
| 82 |
+
]
|
| 83 |
+
}
|
| 84 |
+
```
|
| 85 |
+
|
| 86 |
+
---
|
| 87 |
+
|
| 88 |
+
## Evaluation & Scoring
|
| 89 |
+
|
| 90 |
+
The official contest defines the score as:
|
| 91 |
+
|
| 92 |
+
```
|
| 93 |
+
Score = max(0, β (Precision + Recall β 4 Γ HeightError) Γ 5 Γ 10β΄ β)
|
| 94 |
+
```
|
| 95 |
+
|
| 96 |
+
* **Precision**: Correct predicted building area Γ· total predicted area
|
| 97 |
+
* **Recall**: Correctly matched ground truth area Γ· total ground truth area
|
| 98 |
+
* **HeightError**: Weighted RMSE of predicted vs. true heights for matched buildings
|
| 99 |
+
|
| 100 |
+
**Important constraints:**
|
| 101 |
+
|
| 102 |
+
* β€ 1,000 buildings per image
|
| 103 |
+
* β€ 300 vertices per polygon
|
| 104 |
+
* Total vertices per JSON β€ 5,000
|
| 105 |
+
* Height β [0, 1000] meters
|
| 106 |
+
* Coordinates within [0, 512]
|
| 107 |
+
* No self-intersecting polygons
|
| 108 |
+
* Overlap between two buildings β€ 10% of smaller area
|
| 109 |
+
|
| 110 |
+
---
|
| 111 |
+
|
| 112 |
+
## Usage
|
| 113 |
+
|
| 114 |
+
```python
|
| 115 |
+
from datasets import load_dataset
|
| 116 |
+
|
| 117 |
+
ds = load_dataset("MElHuseyni/building_height_estimation")
|
| 118 |
+
|
| 119 |
+
sample = ds["train"][0]
|
| 120 |
+
print(sample["image"])
|
| 121 |
+
print(sample["buildings"])
|
| 122 |
+
```
|
| 123 |
+
|
| 124 |
+
Each `building` entry contains a list of polygon points and a height value.
|
| 125 |
+
|
| 126 |
+
---
|
| 127 |
+
|
| 128 |
+
## Limitations
|
| 129 |
+
|
| 130 |
+
* **Monocular Input**: Heights are inferred from a single RGB image, no stereo or LiDAR.
|
| 131 |
+
* **Annotation Noise**: Minor misalignments or errors in footprints may exist.
|
| 132 |
+
* **Imbalance**: Height distribution may be skewed (low-rise dominant).
|
| 133 |
+
* **Test Labels**: Hidden; only evaluable via AlgoTester scoring scripts.
|
| 134 |
+
|
| 135 |
+
---
|
| 136 |
+
|
| 137 |
+
## Citation
|
| 138 |
+
|
| 139 |
+
If you use this dataset, please cite both the Hugging Face dataset and the original AlgoTester contest:
|
| 140 |
+
|
| 141 |
+
```
|
| 142 |
+
@misc{building_height_estimation_MElHuseyni,
|
| 143 |
+
title = {building_height_estimation},
|
| 144 |
+
author = {MElHuseyni},
|
| 145 |
+
year = {2025},
|
| 146 |
+
howpublished = {Hugging Face Dataset},
|
| 147 |
+
url = {https://huggingface.co/datasets/MElHuseyni/building_height_estimation}
|
| 148 |
+
}
|
| 149 |
+
|
| 150 |
+
@online{algotester_building_height,
|
| 151 |
+
title = {Building Contour Detection & Height Estimation Contest},
|
| 152 |
+
author = {AlgoTester},
|
| 153 |
+
year = {2024},
|
| 154 |
+
url = {https://algotester.com/en/ContestProblem/DisplayWithFile/135254}
|
| 155 |
+
}
|
| 156 |
+
```
|
| 157 |
+
|
| 158 |
+
---
|
| 159 |
+
|
| 160 |
+
## Acknowledgements
|
| 161 |
+
|
| 162 |
+
* Dataset originally prepared and hosted by **AlgoTester** for their contest.
|
| 163 |
+
* Curated and published on Hugging Face by **[MElHuseyni](https://huggingface.co/MElHuseyni)**.
|
| 164 |
+
* Licensed under MIT for research and development purposes.
|
| 165 |
+
|
| 166 |
+
---
|
| 167 |
+
|