robot

Microsoft Florence-2 Image Annotation Model Open Sourced

Florence-2 is an advanced visual foundation model that employs a prompt-based approach to handle a variety of visual and vision-language tasks. Florence-2 can interpret simple text prompts to perform tasks such as captioning, object detection, and segmentation. It leverages our FLD-5B dataset, which contains 5.4 billion annotations across 126 million images, to master multi-task learning. The model's sequence-to-sequence architecture allows it to excel in zero-shot and fine-tuning settings, proving it to be a competitive visual foundation model.

article image

Florence-2 is a highly advanced visual foundation model that employs an innovative prompt-based approach to handle a wide range of visual and vision-language tasks. This prompt-based method endows Florence-2 with a unique capability to interpret simple text prompts and execute various complex tasks such as caption generation, object detection, and segmentation based on these prompts.

Florence-2 makes full use of our meticulously constructed FLD-5B dataset. This dataset is vast, encompassing up to 126 million images with as many as 5.4 billion annotations. By learning from and mining this rich dataset, Florence-2 has mastered the skills and capabilities of multi-task learning.

The sequence-to-sequence architecture adopted by the model is one of the key factors behind its outstanding performance. This architecture enables Florence-2 to demonstrate exceptional performance in both zero-shot and fine-tuning settings. These excellent performances fully prove that Florence-2 is a highly competitive visual foundation model with significant status and value in the field of vision.