1 paper · 1 filter
Dan Oneata, Herman Kamper
Visually grounded speech models link speech to images. We extend this connection by linking images to text via an existing image captioning system, and as a result gain the ability…