GT-RAIN

GT-RAIN挑战赛邀请公众推动单一图像脱轨的边界，以挑战从世界各地收集到的各种程度的阴雨天气所导致的现实世界图像，这些图像从北美延伸到亚洲。比赛的特点是第一个大规模数据集，包括从115多个场景中捕获的真实多雨图像和地面真相图像对。

The GT-RAIN Challenge invites the public to advance the state-of-the-art of single-image deraining, utilizing real-world images degraded by varying levels of rainy weather collected from across the globe, spanning from North America to Asia. This competition features the first large-scale dataset, which comprises paired real rainy images and their corresponding ground-truth images captured from over 115 distinct scenes.

搜集汇总

背景与挑战

背景概述

GT-RAIN是一个用于图像去雨任务的大规模数据集，包含从115多个全球场景收集的真实多雨图像及其对应的地面真相图像对，旨在推动单一图像脱轨技术边界。该数据集由加州大学、耶鲁大学和DEVCOM Army Research Laboratory于2022年发布，具有现实世界天气多样性特点。