New MIT system extracts hidden videos from visible footage by analyzing how light reflects on a pile of clutter

MIT CSAIL researchers have found a way to reconstruct hidden videos that are being displayed outside a camera's field of view by analyzing how reflections and shadows change in scenes that are visible to the naked eye. Once refined, this method could help self-driving cars better detect obstacles.

You can find the paper, entitled "Computational Mirrors: Blind Inverse Light Transport by Deep Matrix Factorization," on arXiv.org. Read more...

More about Mashable Video, Artificial Intelligence, Mit, Neural Networks, and Tech

from Mashable https://ift.tt/2Pl2TLt