QuPath m5 - convert Pixelclassifiermap into Annotations/Detections does not work

Image Analysis
#1

Hi Pete,

I had a look at Qm5. QuPath is on a great track and I see how valuable the pixel classifier will probably be for Deep Learning. Thank you for the improvements in QuPath!
I experimented with the pixel classifier and tried to convert the segmentation map into annotations or detections on a very small image patch. QuPath starts computing, uses up to 13 GB RAM even it is only 40x40 mircons and never stops.

Is the conversion function already available? (I hope I do report a relevant/potential bug ^-^)

#2

HI @David_C the function can work, but it doesn’t really like very intricate regions. I’d suggest trying a low-resolution classification and seeing if that works, and then seeing how complicated it can handle before it takes too long to be practical.

There may be substantial progress in the next release, but it would be a Big Change…

Initial tests suggest that changing how the hierarchy operations makes QuPath immediately able to work with more complex annotations much more quickly, while most commands continue to behave much the same as before. But since I only started exploring it last night, don’t know exactly what the implications will be… I’ll explore some more today and tomorrow.

1 Like
#3

ok. good to know. Thank you for your response Pete.