Hi,
I’m using label-studio with the SAM2 model from the ML backend. The ML backend takes the whole image (path of the image is provided) and the prompt from the frontend to create the segmentation. I think the segmentation performance would be much better, if one could zoom in relevant regions of the image and instead of the whole image only these zoomed region was used for segmentation. So here is my question: Is it possible to use information from zooming in the ML backend?
Thanks and regards,
Alex