How to ask about coordinates of object in prompt?
#25
by
ZFTurbo
- opened
For example I ask about coordinates of bounding box of some object on picture. It returns something like [345, 303, 450, 600]. But size of image on input of model is normalized as I understand. So how to correctly use it?
It doesn't support.
Hi,
The coordinates [345, 303, 450, 600] are likely normalized or scaled relative to the model's fixed input size (e.g 1024 X 1024). To correctly use them on your original image, you must first convert them to a [0, 1] range by dividing by the normalization factor (e.g 1000 or the input size dimension). Finally, multiply the [0, 1] normalized values by the respective width and height of your original image to get the actual pixel coordinates.
Thanks.