It could probably be taught to recognize faces. In the end, it might not be any better than manual adjustment, but in real-time. Even though it takes milliseconds to render one image, getting it to run every 16-40 ms for video would likely be impossible as nevcairiel said.
Another of example of how deep learning is coming for everyone's job. Silicon Valley is obsessed with it. They can diagnose cancer in an MRI scan in 10 seconds using current neural networks.
|