swaped coordinates in picasso.gausslq._sum_and_center_of_mass

[TG-fresnel]

Hello,

I started using picasso recently and really enjoy it. While looking at some of it's functions i believe i found a bug in the gaussian least squre module
picasso.gausslq. The function _sum_and_center_of_mass computes the center of mass like this:

    x = 0.0
    y = 0.0
    _sum_ = 0.0
    for i in range(size):
        for j in range(size):
            x += spot[i, j] * i
            y += spot[i, j] * j
            _sum_ += spot[i, j]
    x /= _sum_
    y /= _sum_

If im not mistaken in the rest of the code the first dimension in this case the index i corresponds to the y coordinate. So in the end the x and y coordinates are swapped.

I checked the time it takes to fit a gaussian at the coordinates (3,3) where this have no effect, compared to one a (3,-3) where the inital guess for the fit would not be a the right position. I attach the file.

test_fit_spot.py

the averaget time for the fit at (3,3): 0.1018.ms
the averaget time for the fit at (3,-3): 0.6346ms

I guess the effect on real data is smaller since it's usually well centered.

Thanks for the great work with the picasso software and best regards

Tobias

[rafalkowalewski1]

Hi Tobias, thanks for raising the issue. I will look into it in the coming days.

Best,
Rafal

[rafalkowalewski1]

Hi Tobias,

Thanks for raising the issue again, you're completely right. I implemented your suggestion (also in the script gaussmle.py) and it will be available in Picasso 0.9.7 in the next few days.