no code implementations • 13 Apr 2020 • Haoran Li, Weihong Quan, Meijun Yan, Jin Zhang, Xiaoli Gong, Jin Zhou
However, due to the variety of image degradation types in the real world, models trained on single-modal simulation datasets do not always have good robustness and generalization ability in different degradation scenarios.