The training dataset for custom DECODE model #238
Unanswered
Hankeun-Lee
asked this question in
General
Replies: 1 comment 1 reply
-
Dear @Hankeun-Lee,
Best, |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi, I am a researcher at an university and first of all, thank you for your outstanding works!
I am very interested in applying DECODE model to one of our problems, which is to localize nanoparticles from an image acquired by a microscope we developed.
Since the dataset is different from that of yours, I decided to rebuilt the model following your code.
In our dataset, we do not have frames, but rather single images at a given time.
So, I decided to apply a single UNet with multi-head outputs, including intensity, background, prediction probability, and the uncertainties.
It seems like the output image of the uncertainties and prediction probability closely follow the output of your original model.
However, I am having trouble getting the right intensity output.
In my dataset, the intensity of a PSF ranges from 2 to 4, but I can see that you applied sigmoid to the intensity output, which forces the values between 0 to 1.
So my questions are
Any valuable insight is highly appreciated. Thank you in advance for your time.
Beta Was this translation helpful? Give feedback.
All reactions