What is the one thing most photographers want? Maybe seen in the dark to get a low light picture with less noise. It may take some time to achieve this with hardware, but Google‘Research Division’ open source project, MultiNERFCan replace low-light photography sooner than we expected.
Google Research’s RAWNeRF, a part of MultiNeRF research, promises results that put it light years ahead of any other noise-reduction tool. ronerf The tool can read images and, using artificial intelligence, add a high level of detail to photos taken in dark and low-light conditions.
“When adapted to multiple noises raw Input, NERF makes a visual representation so accurate that its presented novel scenes employ dedicated single and multi-image deep raw denoisers over the same wide baseline input images,” reads a Cornell University paper.
What is NERF?
NERF is a neural network tool capable of recreating accurate 3D renders from a set of images. According to one of the Google researchers, Ben MildenhallNERF is designed to work best with well-lit scenarios.
How does MultiNERF improve upon the shortcomings of NERF?
However, when tried with images taken in low light conditions, the results tend to be noisy and compromise details. The problem can be solved with a denoising tool, which can result in further loss of detail.
Meanwhile, in RAWNeRF, algorithms are run on RAW images, and AI is tasked with reducing the noise captured by the sensor while maintaining the detail, let’s see, in the dark.
Google says that RAWNeRF is more capable of reducing noise. It can position the camera to view the scene from different angles, or change the exposure, tone map, and focus “with a precise bokeh effect.”