Florence2Task

Enums · AiDotNet.VisionLanguage.Encoders

The Florence-2 tasks. Each is a task prompt the language encoder reads after the image tokens.

How It Works

The prompts are the ones the Florence-2 processor substitutes for its task tokens (HF Florence2Processor.task_prompts_without_inputs and task_prompts_with_input). Tasks marked “with input” take a phrase or region description; region inputs are written as location tokens.

Fields

Field Summary
Florence2Task Caption
: “What does the image describe?”
Florence2Task CaptionToPhraseGrounding <CAPTION_TO_PHRASE_GROUNDING> (with input): “Locate the phrases in the caption: {input}”
Florence2Task DenseRegionCaption <DENSE_REGION_CAPTION>: “Locate the objects in the image, with their descriptions.”
Florence2Task DetailedCaption <DETAILED_CAPTION>: “Describe in detail what is shown in the image.”
Florence2Task MoreDetailedCaption <MORE_DETAILED_CAPTION>: “Describe with a paragraph what is shown in the image.”
Florence2Task ObjectDetection : “Locate the objects with category name in the image.”
Florence2Task Ocr : “What is the text in the image?”
Florence2Task OcrWithRegion <OCR_WITH_REGION>: “What is the text in the image, with regions?”
Florence2Task OpenVocabularyDetection <OPEN_VOCABULARY_DETECTION> (with input): “Locate {input} in the image.”
Florence2Task RegionProposal <REGION_PROPOSAL>: “Locate the region proposals in the image.”
Florence2Task RegionToCategory <REGION_TO_CATEGORY> (with input): “What is the region {input}?”
Florence2Task RegionToDescription <REGION_TO_DESCRIPTION> (with input): “What does the region {input} describe?”
Florence2Task RegionToOcr <REGION_TO_OCR> (with input): “What text is in the region {input}?”
int value__