Spaces:

gokaygokay

/

Florence-2

like 671

Running on Zero

Unlike PaliGemma, Florence-2 segments only one object per image. For example: For the below image with prompt type REFERENCE_EXPRESSION_SEGMENTATION and text prompt: "Person and car", the output was:

Input image:

Output image:

Only the person was segmented. Is there a way to achieve multiple object segmentation?