---
pretty_name: AeroIntent - City
task_categories:
- video-classification
- object-detection
language:
- en
tags:
- uav
- multimodal
- intent-recognition
- synthetic-data
- unreal-engine
- airsim
size_categories:
- 100K<n<1M
---

# AeroIntent: City Scene

This repository is one scene of **AeroIntent**, a synchronized multimodal benchmark for UAV intent recognition.

## Review Status and Access

AeroIntent is currently under peer review. Access is manually gated during the review period, and the complete benchmark will be made public after the paper is accepted. Please do not redistribute the data or use it for publication before the public release.

## Contents

- 15 UAV configurations from three airframe families and five payload configurations
- 22 trajectory programs grouped into Transit, Reconnaissance, Probing, Strike, and Escape
- Front view only
- RGB, infrared, depth, Echo radar, IMU, GPS, bounding boxes, timestamps, and metadata
- 330 synchronized sequences in this scene

The data are stored as one 	ar.zst shard per UAV configuration. For reliable public transfer, each shard is split into ordered 256-MiB parts. Reconstruct and verify a shard, then extract all shards into the same destination directory.

`ash
python reconstruct_shard.py data/Drone_no_payload
tar --use-compress-program=unzstd -xf Drone_no_payload.tar.zst
`

Each UAV directory contains a manifest.json with the original archive checksum and ordered part checksums. SHA256SUMS.txt lists all uploaded parts. Collection logs, batch logs, completion markers, and non-front views are intentionally excluded.

## Dataset Structure

`	ext
<UAV configuration>/<intent>/<trajectory>/view_front/
  rgb/ ir/ depth/ depth_preview/
  echo/ echo_groundtruth/ imu/ gps/
  labels_all.txt timestamps.csv meta.json
  rgb_video.mp4 ir_video.mp4 depth_video.mp4
`

## Citation and License

Citation and license information will be finalized with the archival release.