UC Berkeley And Adobe AI Researchers Propose BlobGAN, A New Unsupervised And Mid-Level Representation For Insane Scene Manipulation

Since the advent of computer vision, one of the fundamental questions of the research community has always been how to represent the incredible richness of the visual world. One concept that emerged since the beginning is the importance of a *scene* in the context of understanding objects. Suppose we want a classifier for distinguishing between a couch and a bed. In that case, the scene context will give information concerning the surrounding (i.e., the room is a living room or …

adobe ai artificial representation researchers uc berkeley unsupervised