It’s called “Building High-level Features Using Large ... of a human face in a photo—and reliably identify them when they appear. But such training typically requires images that are labeled ...