---
pretty_name: OpenFlyScan
license: other
license_name: openflyscan-component-licenses
license_link: https://huggingface.co/datasets/mistletoe235/openflyscan/blob/main/LICENSES/README.md
language:
  - en
  - zh
tags:
  - robotics
  - gaussian-splatting
  - aerial-reconstruction
  - unreal-engine
  - hardware-in-the-loop
  - arxiv:2609.24253
---

# OpenFlyScan

[Paper](https://arxiv.org/abs/2609.24253) · [Project page](https://openflyscan.github.io/) · [Citation](#citation)

Resources for aerial Gaussian-splat reconstruction and drone simulation.
This dataset repository hosts downloadable runtime packages and their associated
scene resources. Training/evaluation code and the trained Quality Predictor
head are in the [main GitHub repository](https://github.com/mistletoe235/OpenFlyScan).

## Available now

| Resource | Contents | Entry point |
| --- | --- | --- |
| **HIL-simulator v0.1.0 — Expo East** | Linux runtime, preconverted Expo East scene, English HIL monitor and PLY converter | [Download and run](HIL-simulator/README.md) |

The Expo East package opens directly with the **HIL monitor visible**. Without
a connected phone, it remains in the waiting state; an optional software-only
preview is also available. No UE editor or scene conversion is required to
open the included scene.

![Expo East HIL window](previews/expo_east_hil_window.png)

## Repository layout

```text
HIL-simulator/
  README.md
  linux/v0.1.0/
    OpenFlyScan-HIL-ExpoEast-Linux-v0.1.0.tar.zst
    SHA256SUMS
    release_manifest.json
docs/
  simulator_quickstart.md
  hil_quickstart.md
  hil_quickstart_zh.md
manifests/
  scenes.json
previews/
  expo_east_start.png
  expo_east_hil_window.png
  expo_east_english_panels.png
LICENSES/
  README.md
  runtime/
```

This dataset currently contains the simulator package and converted Expo East
scene resources. Standalone GS PLYs and training teacher/cache bundles are not
yet available here. The trained head is already included in the main repository
at `weights/quality_predictor.pt`, together with a small real-feature inference
example; it is not a separate HF download.

## Code and companion apps

- [OpenFlyScan](https://github.com/mistletoe235/OpenFlyScan): trained head, CPU
  example, training/evaluation and the mobile-facing workstation service.
- [OpenFlyScan-UE](https://github.com/mistletoe235/OpenFlyScan-UE): UE source,
  scene conversion and HIL integration.
- [App downloads](https://github.com/mistletoe235/OpenFlyScan/releases/tag/preview-20260922):
  signed Android V4/V5 APKs and iOS access information.
- [TestFlight requests](https://github.com/mistletoe235/OpenFlyScan/discussions/1):
  iOS invitations when an eligible external-testing build is available.

The source repositories and simulator downloads are publicly available.

## Licensing

The Expo East scene resources use **CC BY 4.0**. Original project code uses
Apache-2.0; NanoGaussianSplatting and other dependencies retain their own terms,
and Unreal Engine remains under Epic's EULA. See [license scope](LICENSES/README.md).

The existing v0.1.0 runtime archive and checksum are unchanged. Updated notices,
the scene license and [Eigen source](HIL-simulator/linux/v0.1.0/THIRD_PARTY_SOURCES/README.md)
are provided alongside it; no replacement runtime download is needed.

## Citation

If you use OpenFlyScan or these resources in your research, please cite:

```bibtex
@misc{you2026openflyscanqualityguidedaerialreconstruction,
  title = {OpenFlyScan: A Quality-Guided Aerial Reconstruction System for Consumer Drones},
  author = {Zhongrui You and Zhen Li and Junli Liu and Zhigang Wang and Bin Zhao},
  year = {2026},
  eprint = {2609.24253},
  archivePrefix = {arXiv},
  primaryClass = {cs.RO},
  url = {https://arxiv.org/abs/2609.24253}
}
```