Single image deraining
So, let me consider this task.
What is about deraining? First thought is that this is not a natural task per se. Like, we do not expect a person to derain an image. Sure, there is the rain and the background scene is not so visible.
For safety, we can consider the importance in autonomous car. It would be a system that is not affected in terms of visibility when it is raining. So, the importance of deraining.
Consegui pensar em outro caso. Para surveillance, a camera captured a person in a risk situation with heavy rain, and, even the person detection is hard and can help to identify person in risk.
So, important motivations for deraining.
But also we can think that this is not about single image, because the cases are about videos, but, if we consider the frames, that is the case. The temporal information can help in the rain learning, perhaps, but with a single image we can solve that.
Despite of the motivations, single image deraining segue sem demonstrar uma atenção grande à essas questões. Em uma época onde GenAI is so strong, talvez com conditionals, it can simply generate the derained image, but, for downstream tasks like autonomous driving and object detection, it is not reliable, ainda mais quando se considera imagens com low-quality. Bom, RIS and RID are not that used. This would be a strong case for research. The importance is in there for the learning.