Text-Only Training for Image Captioning using Noise-Injected CLIP.

David Nukrai, Ron Mokady, Amir Globerson

Browse the full EMNLP paper archive.