Incongruency using the mAP function.
GaloIglesias opened this issue · comments
Hello. I am currently using darknet for one of my Computer Vision projects and I have come up with something that does not make much sense. When I run ! ./darknet detector map data/obj.data cfg/yolo-obj.cfg backup/yolo-obj_best.weights to test the model, I get the following output:
How can I get a precision of 92.7% with that distribution of FP (102) and TP (39)? Is there something I am not seeing.
Thanks in advance for the help!
Galo.