SyMTRS: Synthetic Multi-Task Benchmark for Aerial Depth, Domain Adaptation, and Super-Resolution

A new synthetic dataset called SyMTRS has just been created through an innovative urban simulation process. It includes high-resolution RGB aerial images sized at 2048 x 2048, accurate depth maps, nighttime images to help with domain adaptation, and lower-resolution versions for enhancing quality at scales of x2, x4, and x8. Unlike other datasets that focus on just one aspect, SyMTRS acts as a versatile benchmark, enabling joint work in geometry, radiometry, and other domain tasks. It tackles issues like precise depth annotations, control over lighting differences, and multi-scale imagery that have previously slowed progress in monocular depth estimation and super-resolution for aerial images. You can find this research on arXiv with ID 2604.21801.