Paper summary
Paper summary: Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
Florence-2 is a single sequence-to-sequence vision model that takes a text prompt as the task instruction and answers in text, covering captioning, detection, grounding and segmentation, trained on 5.4B annotations.
- Every vision task, including boxes and masks, is turned into "prompt in, text out".
- The data engine (FLD-5B: 5.4B annotations on 126M images) is as much the contribution as the model.