RealRain-1k
收藏资源简介:
RealRain-1k是由IEEE成员创建的高质量真实雨天图像数据集,包含1120对高分辨率清洁和雨天图像,涵盖低密度和高密度雨流。该数据集通过从大量真实世界雨天视频片段中自动生成,采用可控雨密度过滤方法,确保图像具有高分辨率、背景多样性、雨流多样性和严格的空间对齐。RealRain-1k不仅用于单图像去雨研究,还衍生出名为SynRain-13k的大规模合成数据集,通过将真实雨流层粘贴到自然图像上,减少合成与真实图像之间的领域差距。该数据集广泛应用于图像恢复领域,旨在解决真实场景中单图像去雨的挑战。
RealRain-1k is a high-quality real-world rainy image dataset created by IEEE members, consisting of 1120 pairs of high-resolution clean and rainy images covering both low-density and high-density rain streaks. This dataset is automatically generated from a large number of real-world rainy video clips, adopting a controllable rain density filtering method to ensure that the images have high resolution, diverse backgrounds, diverse rain streaks, and strict spatial alignment. RealRain-1k is not only used for single-image deraining research, but also spawned a large-scale synthetic dataset named SynRain-13k, which reduces the domain gap between synthetic and real-world images by pasting real rain streak layers onto natural images. This dataset is widely applied in the field of image restoration, aiming to address the challenges of single-image deraining in real-world scenarios.

- 1Toward Real-world Single Image Deraining: A New Benchmark and BeyondIEEE · 2022年



