MIT Invisible Objects

We humans aren’t very good at spotting objects in the dark. Especially, when it is pitch black where objects appear almost “invisible.”

So a bunch of MIT scientists has developed a deep learning technique that can spot objects in “almost pitch-black conditions”. To make it work; the team developed a neural network to look for transparent patterns in dark images, by feeding it 10,000 dark, grainy and blurred images.

The laws of physics say that light needs to be present in order to fully identify an object. In this case, the developed AI technique highlighted hidden transparent objects by producing ripples in what little light was present. To tackle the blurred part of the images, researchers set their camera to take images slightly out of focus.

The researchers also mentioned in the study that the reconstructed pattern of the image taken in the pitch black was sharper and defined than the reconstruction of the image taken under a light source; 1,000 times brighter to be exact.

Of course, the new AI method will bring fruitful results in the photography. But the researchers are looking forward to using the technology in the medical field.

Minute particles in any biological matter tend to change their properties and structure when exposed to light. This deep learning technique could be used to avoid such changes. It may even help in making groundbreaking discoveries in other fields.

Also Read: This MIT Developed 3D Printer Is 10 Times Faster Than Modern 3D Printers