sgrvinod / a-PyTorch-Tutorial-to-Object-Detection

SSD: Single Shot MultiBox Detector | a PyTorch Tutorial to Object Detection
MIT License
3.04k stars 718 forks source link

What could be causing the AP to come out negative? #72

Open ZoltanV-V opened 3 years ago

ZoltanV-V commented 3 years ago

Hi there, I am training SSD for the first time using an XView dataset but during the evaluation step, the class average precision comes out as negative? Here's what it shows for a couple of the classes:

AP for Fixed-wing Aircraft = -1.0000 AP for Small Aircraft = -1.0000 AP for Car = -1.0000

The mAP is 77.2. I am not sure why the AP is showing as a negative? I was expecting it to be in the range of [0,1] but maybe I am misunderstanding something. Any tips would be helpful, thank you!