---
pretty_name: "SegFly: A Dataset and 2D-3D-2D Paradigm for Aerial RGB-Thermal Semantic Segmentation at Scale"
task_categories:
  - image-segmentation
task_ids:
  - semantic-segmentation
license: cc-by-nc-sa-4.0
language:
  - en
size_categories:
  - 10K<n<100K
tags:
  - segmentation
  - semantic-segmentation
  - multimodal
  - aerial
  - drone
  - uav
  - remote-sensing
  - rgb-thermal
  - thermal
---

<div align="center">

# SegFly: A Dataset and 2D-3D-2D Paradigm for Aerial RGB-Thermal Semantic Segmentation at Scale


[![GitHub Documentation](https://img.shields.io/badge/GitHub-Documentation-green?logo=github&logoColor=white&labelColor=555)](https://github.com/markus-42/SegFly)&nbsp;&nbsp;
[![Project](https://img.shields.io/badge/Project-Page-blue.svg)](https://markus-42.github.io/publications/2026/segfly/)&nbsp;&nbsp;
[![arXiv](https://img.shields.io/badge/arXiv-Paper-red.svg)](https://arxiv.org/abs/2603.17920)&nbsp;&nbsp;
[![Firefly Model](https://img.shields.io/badge/HuggingFace-Firefly%20Model%20-e58f00?logo=huggingface&logoColor=white&labelColor=555)](https://huggingface.co/markus-42/SegFly-Firefly)



</div>

Following the acceptance of [SegFly](https://markus-42.github.io/publications/2026/segfly/) at the ECCV 2026 computer vision conference, we officially release our dataset on Hugging Face. Note that we also release our [Firefly](https://huggingface.co/markus-42/SegFly-Firefly) model, specialized for RGB and thermal semantic segmentation from aerial perspectives.

SegFly is a large-scale aerial semantic segmentation dataset featuring 20,606 high-resolution RGB images and 15,007 pixel-aligned RGB-Thermal (RGB-T) pairs. Images are captured across urban, industrial, and rural environments, spanning all seasons and three altitudes (30m, 40m, 50m).



## Quickstart

> **Full dataset documentation available on [SegFly GitHub](https://github.com/markus-42/SegFly).**

```python
from datasets import load_dataset

# Load entire dataset
dataset = load_dataset("markus-42/SegFly")
```

## Features

| Feature | Type | Description |
| :--- | :---: | :--- |
| `image` | `Image` | Raw sensor frame (RGB or LWIR Thermal) |
| `label` | `Image` | 8-bit single-channel semantic mask mapped to 15 benchmark classes |
| `RGB_aligned` | `Image` | Registered RGB image (Thermal modality only; returns `None` for RGB modality) |
| `scene` | `string` | Scene identifier (`"scene_01"` to `"scene_09"`) |
| `altitude` | `string` | Flight altitude (`"30m"`, `"40m"`, `"50m"`) |
| `modality` | `string` | Sensor modality (`"RGB"` or `"thermal"`) |

## Splits and Statistics

* **Total Samples**: 35,613 (20,606 RGB + 15,007 thermal)

| Modality | Split | Scenes | Sample Count |
| :--- | :--- | :--- | :---: |
| **RGB** | Train | `scene_01`, `scene_02`, `scene_03`, `scene_04`, `scene_05` | 14,738 |
| | Val | `scene_06`, `scene_07` | 1,965 |
| | Test | `scene_08`, `scene_09` | 3,842 |
| **Thermal** | Train | `scene_03`, `scene_04`, `scene_05` | 12,063 |
| | Val/Test | `scene_09` | 2,944 |

## SegFly Dataset Class Mapping Reference

| Class ID | Class Name | RGB Color | Color Preview |
| :---: | :--- | :---: | :---: |
| 0 | Unlabeled / Ignored | `[0, 0, 0]` | <span style="display:inline-block; width:12px; height:12px; background-color:#000000; border:1px solid #000; margin-right:5px;"></span>`#000000` |
| 1 | Road | `[128, 0, 128]` | <span style="display:inline-block; width:12px; height:12px; background-color:#800080; border:1px solid #000; margin-right:5px;"></span>`#800080` |
| 2 | Walkway | `[204, 163, 72]` | <span style="display:inline-block; width:12px; height:12px; background-color:#cca348; border:1px solid #000; margin-right:5px;"></span>`#cca348` |
| 3 | Dirt | `[128, 0, 0]` | <span style="display:inline-block; width:12px; height:12px; background-color:#800000; border:1px solid #000; margin-right:5px;"></span>`#800000` |
| 4 | Gravel | `[192, 192, 192]` | <span style="display:inline-block; width:12px; height:12px; background-color:#c0c0c0; border:1px solid #000; margin-right:5px;"></span>`#c0c0c0` |
| 6 | Grass | `[0, 255, 0]` | <span style="display:inline-block; width:12px; height:12px; background-color:#00ff00; border:1px solid #000; margin-right:5px;"></span>`#00ff00` |
| 7 | Vegetation | `[112, 148, 32]` | <span style="display:inline-block; width:12px; height:12px; background-color:#709420; border:1px solid #000; margin-right:5px;"></span>`#709420` |
| 8 | Tree | `[64, 64, 0]` | <span style="display:inline-block; width:12px; height:12px; background-color:#404000; border:1px solid #000; margin-right:5px;"></span>`#404000` |
| 9 | Ground Obstacle | `[255, 255, 0]` | <span style="display:inline-block; width:12px; height:12px; background-color:#ffff00; border:1px solid #000; margin-right:5px;"></span>`#ffff00` |
| 13 | Vehicle | `[0, 128, 128]` | <span style="display:inline-block; width:12px; height:12px; background-color:#008080; border:1px solid #000; margin-right:5px;"></span>`#008080` |
| 14 | Water | `[0, 0, 255]` | <span style="display:inline-block; width:12px; height:12px; background-color:#0000ff; border:1px solid #000; margin-right:5px;"></span>`#0000ff` |
| 16 | Building | `[255, 0, 0]` | <span style="display:inline-block; width:12px; height:12px; background-color:#ff0000; border:1px solid #000; margin-right:5px;"></span>`#ff0000` |
| 17 | Roof | `[64, 160, 120]` | <span style="display:inline-block; width:12px; height:12px; background-color:#40a078; border:1px solid #000; margin-right:5px;"></span>`#40a078` |
| 33 | Parking Lot | `[128, 64, 128]` | <span style="display:inline-block; width:12px; height:12px; background-color:#804080; border:1px solid #000; margin-right:5px;"></span>`#804080` |
| 34 | Construction | `[240, 120, 120]` | <span style="display:inline-block; width:12px; height:12px; background-color:#f07878; border:1px solid #000; margin-right:5px;"></span>`#f07878` |
| 36 | Truck | `[128, 128, 64]` | <span style="display:inline-block; width:12px; height:12px; background-color:#808040; border:1px solid #000; margin-right:5px;"></span>`#808040` |


## Reference

```bibtex
@inproceedings{gross2026segfly,
    title={{SegFly: A Dataset and 2D-3D-2D Paradigm for Aerial RGB-Thermal Semantic Segmentation at Scale}}, 
    author={Markus Gross and Sai Bharadhwaj Matha and Rui Song and Viswanathan Muthuveerappan and Conrad Christoph and Julius Huber and Daniel Cremers},
    booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},
    year={2026},
}
```

## License
This work is licensed under the [CC BY-NC-SA 4.0 license](https://creativecommons.org/licenses/by-nc-sa/4.0/). See the LICENSE file for the full legal terms.