Spotlight: Sep 18, 2018
Combining recognition of speech and objects, a CSAIL-developed system can learn to identify items in an image based on a spoken description: Given an image and an audio caption, it highlights relevant regions of the image in real time. Full story