Nice tool :-) However, I encountered a problem:
Either my images are super peculiar or the claim that the Optimized Model can be applied to most applications is exaggerated.
This is my result when segmenting with it (Colon tumor organoids):
I followed your instructions step-by-step, but maybe I forgot some variable... Do you know if I could try something out to improve the results or do I need to retrain the model for my data?
It would be useful to provide more information on optimal imaging conditions for the Opitimized Model, such as:
Which magnification?
Which bit-depth?
Color value range in the images?
Which microscope(s) were used for the training data?
In my case I would also appreciate better documentation on how to retrain, including tips on how to create the segmented images:
Recommended tools for annotation? (Labkit?)
How many images needed?
Best image size (e.g. how many organoids per image?)
different conditions necessary (such as density, organoid size)? If yes, how do they need to be distributed?
I'm happy to provide you with images in case you want to test something out :-)
Theresa-S
commented
1 year ago
Hello @jono-m ,
Nice tool :-) However, I encountered a problem: Either my images are super peculiar or the claim that the Optimized Model can be applied to most applications is exaggerated. This is my result when segmenting with it (Colon tumor organoids):
I followed your instructions step-by-step, but maybe I forgot some variable... Do you know if I could try something out to improve the results or do I need to retrain the model for my data?
It would be useful to provide more information on optimal imaging conditions for the Opitimized Model, such as:
In my case I would also appreciate better documentation on how to retrain, including tips on how to create the segmented images:
I'm happy to provide you with images in case you want to test something out :-)
Best, Theresa