experiencor / keras-yolo3

Training and Detecting Objects with YOLO3
MIT License
1.61k stars 861 forks source link

How to get precision and recall values #319

Closed ghost closed 2 years ago

ghost commented 2 years ago

In utils.util, in the evaluate() function, there are two variables precision and recall. However, these are numpy arrays. To get a single value, do you average them all out? I tried averaging them, but then I got a low precision value (0.2) which is confusing since my mAP is 0.65. However, recall seems to be very high at 0.75. It is strange since with other yolov3 implementations the "recall" and "precision" doesn't vary by this much.