Text-only Training For Image Captioning Using Noise-injected CLIP | Awesome LLM Papers Add your paper to Awesome LLM Papers

Text-only Training For Image Captioning Using Noise-injected CLIP

David Nukrai, Ron Mokady, Amir Globerson . Findings of the Association for Computational Linguistics: EMNLP 2022 2022 – 47 citations

[Paper]   Search on Google Scholar   Search on Semantic Scholar
ACL Compositional Generalization EMNLP Has Code Image Text Integration Interdisciplinary Approaches Training Techniques Visual Contextualization

We consider the task of image-captioning using only the CLIP model and additional text data at training time, and no additional captioned images. Our approach relies on the fact that CLIP is trained to make visual and textual embeddings similar. Therefore, we only need to learn how to translate CLIP textual embeddings back into text, and we can learn how to do this by learning a decoder for the frozen CLIP text encoder using only text. We argue that this intuition is “almost correct” because of a gap between the embedding spaces, and propose to rectify this via noise injection during training. We demonstrate the effectiveness of our approach by showing SOTA zero-shot image captioning across four benchmarks, including style transfer. Code, data, and models are available on GitHub.

Similar Work