Flare7K++

Flare7K++是首个针对夜间flare移除的综合性数据集，由南洋理工大学S-Lab创建。该数据集包含962张真实捕获的flare图像（Flare-R）和7000张合成flare图像（Flare7K），旨在解决由于人工光源的独特亮度和光谱以及flare的多样模式和图像退化带来的挑战。Flare7K++通过提供丰富的flare组件注释，如光源、streak、glare等，为夜间flare移除研究提供了宝贵的资源和基准。此外，数据集的创建过程涉及使用不同类型的镜头和光源在暗室中捕获图像，以及基于真实世界夜间flare的观察和统计来合成flare图像。Flare7K++的应用领域包括但不限于夜间摄影、自动驾驶和航空物体跟踪，旨在提高这些领域中视觉算法的性能和可靠性。

Flare7K++ is the first comprehensive dataset dedicated to nighttime flare removal, created by S-Lab at Nanyang Technological University. This dataset includes 962 real-world captured flare images (Flare-R) and 7000 synthetic flare images (Flare7K), aiming to address the challenges brought by the unique brightness and spectrum of artificial light sources, as well as the diverse flare patterns and image degradation. Flare7K++ provides valuable resources and benchmarks for nighttime flare removal research by offering rich annotations of flare components such as light sources, streaks, glares, etc. In addition, the dataset creation process involves capturing images in a darkroom with different types of lenses and light sources, as well as synthesizing flare images based on observations and statistics of real-world nighttime flares. The application fields of Flare7K++ include but are not limited to nighttime photography, autonomous driving and aerial object tracking, with the aim of improving the performance and reliability of visual algorithms in these domains.