The results obtained with my own modified dataset are not desired outcome.

[zhaorongge]

Hello,

A simulation of a dataset is required in which a single two-dimensional coordinate corresponds to multiple photon counts, for example, six. Six frames of an image correspond to a set of coordinates. The class Gauss6PSF (PSF) was added primarily in the file decode.simulation.psf_kernel.py. In this process, scaling is performed on the weight, and frames (n, 6, H, W) are generated from psf.forward. Modifications were made to the relevant calculations in detail. The result of positioning through these modifications is not optimal. The results are presented below. I am seeking assistance in identifying where I may have erred in my approach. Additionally, I would appreciate guidance on how to modify decode's emitterset to better align with my objectives.

Thanks,
Rongge Zhao

[Haydnspass]

Dear @zhaorongge ,
I don't quite get the problem in its entirety. Do you use the GaussianPSF as is, or did you extend/modify it?
By looking at the results, it seems either the x/y extents or the px size is off.
Could you provide us with the modified code so I can have a more in-depth look?

Best,
Lucas