Relonch AI, a California-based startup, recently introduced a new deep learning app that drastically enhances photos shot in low light. The purpose of the application is to introduce users to some of the company’s capabilities in the digital imaging sector.
The app, named Relonch Alfred, corrects photos by creating exposure and color maps of each image. Once a map is created, the system adjusts the lightning in every section of the photograph independently, identifying the background and foreground, and areas that require additional lighting.
“After successful tests of Relonch AI with our Relonch 291 Cameras during 2016-2018, we’re bringing our AI to hundreds of millions of creative and talented smartphone users worldwide,” said Sergey Korzhenevich, co-founder and CEO of Relonch. “The key part of our technology is the intelligence of our AI that adds drama into each story, filling it with an atmosphere the user originally intended, while keeping the image natural looking.”
Using NVIDIA Tesla GPUs and the cuDNN-accelerated TensorFlow deep learning framework, the team trained their system on over 100,000 raw images shot in different lighting conditions. The team also fed the neural network the corresponding edited photos, which they processed manually to mirror artificial light.
Even though the app does not require a dual-lens camera, the company is currently working on an update that will enable smartphone users with dual cameras to enhance their photos, using the camera’s depth maps.
Relonch Alfred is currently available on Apple the App Store, but the team is also working on an Android version.
It’s worth noting, many of the features seen in the videos above, are not yet available on the app. Right now the demo version can only edit existing photos. The company plans to add video compatibility in a future release, as well as the ability to get real-time lighting assistance when taking a photograph.