Researchers at MIT have found a way to “see” invisible objects in total darkness. While deep neural networks have been used in computer vision and image recognition, MIT professor of mechanical engineering George Barbastathis and his team have discovered that it’s possible to “train” a neural network to reconstruct images of objects viewed in total darkness, after feeding the network thousands of images of different objects that would normally be impossible to see in the dark. Once the neural network has been “trained”, it can then recognize and classify totally new images viewed in the dark. To see the full story, read here: http://news.mit.edu/2018/deep-learning-dark-objects-1212

Leave a Reply