How Deep Learning Can Shed Light on Nighttime Video FootageNovember 30, 2017
From selfies to satellites, cameras are an integral part of life. They increasingly watch over our homes and streets, and keep businesses secure inside and out.
But many factors — rain, smog, poor lighting, etc. — can reduce the quality of images. And from identifying a thief to checking on your baby via a monitor, these factors can impair the decisions people make based on camera footage.
NVIDIA Inception partner Brighter AI is focused on a fix. The Berlin-based startup is using deep learning to help people clearly see the images, video footage and streamed pictures our cameras capture, even at night.
Turning Night Into Day
Especially for security cameras, the dark of night is a crucial inhibitor when it comes to capturing clear, sharp, detailed images. Security professionals typically rely on images captured by closed circuit TV or body-worn cameras to determine security threats, and if further action is needed.
Cameras with night vision capabilities are remarkable devices, but they impair decision making given the low levels of detail within captured images. And night vision limits understanding of video footage since the display is monochrome, with little differentiation between objects as well as poor shading.
Simply put, the current standard of cameras don’t give security professionals the required visual support to monitor and react to potential criminal activities.
Brighter AI is using deep learning powered by NVIDIA GPUs to turn images from night vision cameras into full color, defined images. In their own words, they’re “making it daylight again.”
To turn the lights on, Brighter AI uses generative adversarial networks (GANs), a deep learning method that pits.two neural networks, a generator and a discriminator, against each other. The two networks continuously challenge the results of the other to improve their results.
For Brighter AI, the generator network creates daylight images from night vision ones and the discriminator network determines whether or not the daylight image is real or faked. Over time, the networks continue to improve at their task.
Brighter AI’s patent-pending system currently runs up to seven deep neural networks simultaneously for training. It lets professionals base decisions based on video footage with greater confidence, accuracy and speed. Also, since Brighter AI’s solution is a cloud-based API, camera equipment can be upgraded at any time without a hardware change.
Brighter AI is one of more than 2,000 startups in our Inception program. The virtual accelerator program provides startups with access to technology, expertise and marketing support.