Scientists at the Massachusetts Institute of Technology (MIT)'s Computer Science and Artificial Intelligence Laboratory have advanced a technique to envision scenes outside a line of sight without special gear.
The passive method can reconstruct concealed images from subtle shadows and reflections. An algorithm monitors how shadow and geometry interact to predict the path light travels in a scene, then calculates what is unseen from the observed shadows—and can build a silhouette of a live-action performance.
The algorithm concurrently trains two neural networks specialized for a single target: one creates the scrambling pattern, while the other estimates the hidden video.
The networks are rewarded when the blended factors replicate the video recorded from the clutter, driving them to explain their observations with plausible concealed data.
From MIT News
View Full Article
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA
No entries found