The article introduces UAV-Rain1k, a new benchmark dataset for removing raindrops from unmanned aerial vehicle (UAV) images. The researchers created the dataset to address the lack of focus on raindrop removal from UAV aerial imagery, a challenge exacerbated by varying angles and rapid movement during drone flight. They provide a dataset generation pipeline which includes modeling raindrop shapes, collecting background images from various UAV angles, and random sampling of rain masks. The dataset will be publicly available for further research and development in improving the quality of UAV imagery.
Publication date: 9 Feb 2024
Project Page: https://github.com/cschenxiang/UAV-Rain1k
Paper: https://arxiv.org/pdf/2402.05773
