Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
This work addresses the need for more versatile and instruction-following vision models for researchers and practitioners in computer vision, though it is incremental as it builds on existing large vision model paradigms.
The authors tackled the problem of creating a unified vision foundation model that can handle diverse computer vision tasks with simple text prompts, resulting in Florence-2, which demonstrated strong zero-shot and fine-tuning capabilities across numerous tasks.
We introduce Florence-2, a novel vision foundation model with a unified, prompt-based representation for a variety of computer vision and vision-language tasks. While existing large vision models excel in transfer learning, they struggle to perform a diversity of tasks with simple instructions, a capability that implies handling the complexity of various spatial hierarchy and semantic granularity. Florence-2 was designed to take text-prompt as task instructions and generate desirable results in text forms, whether it be captioning, object detection, grounding or segmentation. This multi-task learning setup demands large-scale, high-quality annotated data. To this end, we co-developed FLD-5B that consists of 5.4 billion comprehensive visual annotations on 126 million images, using an iterative strategy of automated image annotation and model refinement. We adopted a sequence-to-sequence structure to train Florence-2 to perform versatile and comprehensive vision tasks. Extensive evaluations on numerous tasks demonstrated Florence-2 to be a strong vision foundation model contender with unprecedented zero-shot and fine-tuning capabilities.