Closed wilmerwang closed 5 years ago
@ilikewind yes, for Normal_*.json files they were generated by random sampling coordinates from the normal tissue mask at some higher level, e.g. level 6.
@yil8 I think the Normal_*.json should include all normal tissue, e.g. Normal_001.json represent the tissue area of Normal_001.tif. so I was confused about random sampling... thanks.
@ilikewind It's always impossible to enumerate all the tissue regions of a normal samples, so you have to do subsampling.
@yil8 In order to check what Normal_.json are, I translate Normal_001.json to Normal001.xml, so I can visualiize it. but it is obvious that normal.json include all normal tissue. It is a intersting to use normal*.json than normal tissue mask images.
and I also want to know more detail you employ for the normal_.json generated. thank you .
@ilikewind I'm not quite sure how you generated this figure, it seems the blue dots are concentrated about boundary regions. This might due to the effect of hard negative mining.
@yil8 the figure was generated by your file Normal-001.json at NCRF/jsons/train/. And I am not sure if you mistake the json file for coordinates/normal_train.txt . It seems we are discuss different issues. Best wish.
@ilikewind I am so sorry, I was completely wrong about your question, and indeed I confused between Normal_001.json with coordinates/normal_train.txt. Normal001.json was not really used through out this code base. Honestly, I forgot how Normal*.json were generated, but they were probably not generated base on tissue mask
@yil8 Ok, thanks for your patient. I will do something about it.
Hi,
I was wondering how to generater Normal_*. json files?
And if it was generatered by the normal tissue mask? can you share the process detail. Thank you!
Kind regards,