LAFITE: Towards Language-Free Training for Text-to-Image Generation
This addresses the high cost and time of human captioning for text-to-image generation, offering a more efficient training approach that could benefit researchers and developers in AI and creative applications.
The paper tackles the challenge of needing large paired image-text datasets for training text-to-image generation models by proposing a method that trains without any text data, leveraging CLIP's multi-modal semantic space to generate text features from images. It achieves state-of-the-art results, outperforming many models trained with full pairs, and demonstrates competitive zero-shot performance on MS-COCO with only about 1% of the model size and training data compared to DALL-E.
One of the major challenges in training text-to-image generation models is the need of a large number of high-quality image-text pairs. While image samples are often easily accessible, the associated text descriptions typically require careful human captioning, which is particularly time- and cost-consuming. In this paper, we propose the first work to train text-to-image generation models without any text data. Our method leverages the well-aligned multi-modal semantic space of the powerful pre-trained CLIP model: the requirement of text-conditioning is seamlessly alleviated via generating text features from image features. Extensive experiments are conducted to illustrate the effectiveness of the proposed method. We obtain state-of-the-art results in the standard text-to-image generation tasks. Importantly, the proposed language-free model outperforms most existing models trained with full image-text pairs. Furthermore, our method can be applied in fine-tuning pre-trained models, which saves both training time and cost in training text-to-image generation models. Our pre-trained model obtains competitive results in zero-shot text-to-image generation on the MS-COCO dataset, yet with around only 1% of the model size and training data size relative to the recently proposed large DALL-E model.