RAW-Bench
收藏资源简介:
RAW-Bench是一个综合性的基准数据集,由东京大学的研究团队创建,旨在评估当前基于RAW的视觉框架在各种退化情况下的性能。该数据集包括17种影响RAW视觉性能的常见退化类型,涵盖了光照退化、天气效应、模糊、相机成像退化以及相机颜色响应的变化。这些退化类型的设计考虑了真实世界中的多种场景,以全面评估模型在领域内和领域外的性能表现。
RAW-Bench is a comprehensive benchmark dataset developed by a research team from The University of Tokyo, which is designed to evaluate the performance of current RAW-based visual frameworks under various degradation scenarios. The dataset includes 17 common degradation types that affect RAW visual performance, covering illumination degradation, weather effects, blurring, camera imaging degradation, and variations in camera color response. The design of these degradation types takes into account multiple real-world scenarios to enable comprehensive evaluation of model performance both in-domain and out-of-domain.




