SyMTRS: Synthetic Multi-Task Aerial Imagery with 1,000 UE5 Records

SyMTRS provides approximately 1,000 synthetic aerial images generated using Unreal Engine 5 by safouaneelg in early 2026. The collection supports multi-task learning by providing high-fidelity simulated environments for depth estimation, domain adaptation, and super-resolution benchmarks.

Use Cases

• Training monocular depth estimation models using synthetic ground truth depth maps
• Benchmarking super-resolution algorithms on high-fidelity simulated aerial imagery
• Developing domain adaptation techniques to bridge the gap between UE5 simulations and real-world remote sensing

Strengths

• High-fidelity simulation using Unreal Engine 5
• Multi-task support for depth, super-resolution, and domain adaptation
• Apache 2.0 license allows for broad research and commercial use

Limitations

• Small scale with approximately 1,000 records
• Synthetic-to-real domain gap inherent in simulated imagery
• Geographic diversity limited to simulated US environments

Provenance

Source

safouaneelg via Hugging Face

Collection Method

Synthetic generation using Unreal Engine 5

Freshness

Last updated February 2026.

Geography

United States (simulated)