Abstract
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 at https://github.com/DavidHuji/CapDec.- Anthology ID:
- 2022.findings-emnlp.299
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2022
- Month:
- December
- Year:
- 2022
- Address:
- Abu Dhabi, United Arab Emirates
- Editors:
- Yoav Goldberg, Zornitsa Kozareva, Yue Zhang
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 4055–4063
- Language:
- URL:
- https://aclanthology.org/2022.findings-emnlp.299
- DOI:
- 10.18653/v1/2022.findings-emnlp.299
- Cite (ACL):
- David Nukrai, Ron Mokady, and Amir Globerson. 2022. Text-Only Training for Image Captioning using Noise-Injected CLIP. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 4055–4063, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
- Cite (Informal):
- Text-Only Training for Image Captioning using Noise-Injected CLIP (Nukrai et al., Findings 2022)
- PDF:
- https://preview.aclanthology.org/proper-vol2-ingestion/2022.findings-emnlp.299.pdf