LSFD
收藏资源简介:
LSFD是一个大规模的真实户外场景去雾数据集,由字节跳动与新加坡国立大学合作创建。该数据集包含10000张图像,其中8000张用于训练,2000张用于测试。数据集通过从高清3D电影中选取高质量的真实户外场景帧,并利用立体深度信息渲染雾霾效果而生成。LSFD数据集的特点在于其高度的真实性和多样性,以及高质量的深度信息,这有助于提升去雾算法在真实场景中的性能。数据集主要应用于提升自动驾驶、安全监控等领域的图像去雾技术。
LSFD is a large-scale real-world outdoor scene dehazing dataset co-created by ByteDance and the National University of Singapore. This dataset consists of 10,000 images, with 8000 allocated for training and 2000 for testing. It is generated by selecting high-quality real outdoor scene frames from high-definition 3D films, and rendering synthetic haze effects using stereo depth information. The LSFD dataset features high realism, diversity, and high-quality depth information, which helps improve the performance of dehazing algorithms in real-world scenarios. This dataset is primarily applied to advance image dehazing technologies in fields such as autonomous driving and security monitoring.

- 1Realistic Large-Scale Fine-Depth Dehazing Dataset from 3D Videos字节跳动与新加坡国立大学 · 2022年



