4 ms·
But wouldn't that level of glare be what would actually happen if you took the original image, and shot it in the amount of light required to make it look like
by derefr 7y ago
But wouldn't that level of glare be what would actually happen if you took the original image, and shot it in the amount of light required to make it look like the output image?
It's not trying to make things readable; it's trying to make things look like there was more light in the room when they were shot. In rooms with high lighting, some objects have glare. That's "correct"—it's what would appear in the training data.