Spectral Use-Case Validation

Super-Resolution: Spectral Validation for Flood 🌊 and Fire 🔥

This interactive visualization evaluates the impact of super-resolution on downstream spectral analysis tasks using Sentinel-2 imagery. The results are generated with the OpenSR latent diffusion model in combination with SEN2SR , a spectrally consistent super-resolution framework designed for Earth-observation data.

The maps compare low-resolution inputs against super-resolved outputs for two real-world use cases: flood-water delineation and burn-scar mapping. Rather than emphasizing visual sharpness alone, the focus is on how improved spatial detail influences the structure and coherence of spectral index responses used in operational remote-sensing workflows.

The full implementation, data processing pipeline, and evaluation code are available in the OpenSR spectral use-cases repository .

Key takeaway Spectrally constrained super-resolution leads to more spatially coherent and better-defined index responses. Improvements are most visible along object boundaries and fine structures, where super-resolution sharpens transitions without distorting the underlying spectral signal.

What to explore Toggle between low-resolution and super-resolved layers and examine how spatial detail changes the delineation of flood extents and burn scars. Pay particular attention to boundary regions, thin structures, and fragmented areas, where super-resolution produces more continuous and geometrically consistent patterns compared to the low-resolution inputs.

🌊 Flood Map 🔥 Burn Scar Map

Flood Map

Burn Scar Map