Low-Altitude-Drone-Remote-Sensing-Dataset

Dataset Description

The Low-Altitude-Drone-Remote-Sensing-Dataset is a high-resolution drone image dataset collected at a consistent low altitude over residential neighborhoods in coastal urban areas, Galveston, Texas. The dataset is designed for remote sensing image compression, image restoration, and downstream visual understanding tasks.

Compared with many existing UAV datasets captured under highly diverse viewpoints and altitudes, this dataset focuses on:

• consistent low-altitude acquisition (approximately 47 m),
• nadir or near-nadir viewpoints,
• high-resolution imagery with rich structural details,
• residential coastal urban scenes containing buildings, streets, parking areas, and vehicles.

Supported Tasks

• Remote sensing image compression
• Perceptual image reconstruction / restoration
• Object detection
• Benchmarking downstream task robustness under compression

Data

• Image format: JPG RGB drone image
• Width: 5472
• Height: 3648

Example Image

Example high-resolution drone image from the dataset.

Uses

Direct Use

This dataset is intended for:

• evaluation of remote sensing image compression methods,
• perceptual reconstruction research,
• object detection on compressed or reconstructed UAV imagery,
• benchmarking task-aware compression performance.

Licensing Information

This dataset is collected by Disaster Data Reconnaissance Center (DDRC) at the Institute for a Disaster Resilient Texas (IDRT), TAMU, USA.

Citation

If you use this dataset, please cite:

@misc{han2026ppobasedbitrateallocationconditional,
      title={A PPO-Based Bitrate Allocation Conditional Diffusion Model for Remote Sensing Image Compression}, 
      author={Yuming Han and Jooho Kim and Anish Shakya},
      year={2026},
      eprint={2603.15365},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2603.15365}, 
}