Inference error

[Zhangyuge721]

Dear Author,

I apologize for the inconvenience, but I encountered two issues after training my network with my own dataset:

There are many missed cell detections.
There are only two types of cells in total, so I set nr_types to 3 (including the background). However, the inference results show three classes, and I did not set a black overlay.
I hope you can assist me with these issues. Thank you!
type_info.json设置

{
"0" : ["nolabe", [0 , 0, 0]],
"1" : ["Normal", [255, 0, 0]],
"2" : ["Cancer", [0 , 255, 0]]
}

[simongraham]

Black overlay means there is a nucleus segmented by the NP branch, but there are no pixels predicted from the NC branch