Scientists from the Inception Institute of Artificial Intelligence (national research organisation based in the UAE), Beijing Institute of Technology, and Stony Brook University (New York) have developed a new form of artificial intelligence that removes the blur in photos.
The model distinguishes between objects and faces of people. To train the algorithm, a data set of several tens of thousands of photos was collected. The Human-aware Image Deblurring (HIDE) set consisted of paired outdoor photographs. On one pair, the photos on the left were blurred, on the right — clear. Pictures were recognized using a human detection model. Scientists used the video card Nvidia Titan X.
“Our model is capable of reconstructing an image with an explicit structure and semantic details.” According to them, the trained model achieves modern characteristics.