microsoft

/

Florence-2-large

like 1.1k

hi @lucasjin , the output coordinates is not 0-1, but within image width and height, please checkout this notebook for more details https://huggingface.co/microsoft/Florence-2-large/blob/main/sample_inference.ipynb

does the input resized?

hi @lucasjin , yes, image input to model will be resized to square.

Without any pad?
if so, the wil lthe coodrinate calculate for both in and out

hi @lucasjin , the output coordinates is with respect to the original image. you don't need to consider image transforms for the inference output format.

but if visualize on oroginal image, the input image are procesed (like resized), does the output coordnates should transform back?