Paper summary
Paper summary: A Picture is Worth a Thousand Words: Principled Recaptioning Improves Image Generation
RECAP relabels a text-to-image training set with captions from a specialised captioning model instead of web alt-text, and reports better image quality and much better prompt following.
- Web alt-text is a poor description of the image; better captions alone improve a text-to-image model.
- The gains show up in semantic accuracy, counting and positional alignment, not only in FID.