FReSH-IR❄️🌧️

Abstract: Adverse weather conditions such as rain, haze, and snow degrade image quality. FReSH-IR is a lightweight all-in-one image restoration method that processes high- and low-frequency information at multiple scales to restore images affected by different weather degradations, aiming for a strong balance between restoration quality and computational efficiency.

Paula Garrido-Mellado, Daniel Feijoo, Yuning Cui, Álvaro García, Marcos V. Conde Fundación Cidaut

Available code at github. More information on the Arxiv paper.

Disclaimer: this is a research demo — results may vary on images with extreme degradations or resolutions very different from the training distribution. Upload an image affected by raindrops, rain, haze, or snow.