Coordinate Heatmap

Hi, Thank you very much for reading my question. as stated in title, if I am trying to regress a coordinate in a 512x512 image ;

oo = np.zeros((512, 512))
oo[212,313] = 1   <-- target

how can I convert this point to a ‘bigger target’ by implementing gaussian range ? and would it be beneficial ? also, what would the final layer of my model be like and what loss function should I be using ?

Thank you very much for your time in reading and helping my with my question and sorry for asking so many questions at once . Thank you !