predict.py预测单张图片准确度很高（目标全都能框对），但是get_map.py时候，mAP只有5%，为啥嘞

[913200451]

0.09% = ClassA AP || score_threhold=0.5 : F1=0.00 ; Recall=60.78% ; Precision=0.10%
3.10% = ClassB AP || score_threhold=0.5 : F1=0.08 ; Recall=65.17% ; Precision=4.33%
3.33% = ClassC AP || score_threhold=0.5 : F1=0.09 ; Recall=63.29% ; Precision=4.78%

[Lrangeo]

0.09% = ClassA AP || score_threhold=0.5 : F1=0.00 ; Recall=60.78% ; Precision=0.10% 3.10% = ClassB AP || score_threhold=0.5 : F1=0.08 ; Recall=65.17% ; Precision=4.33% 3.33% = ClassC AP || score_threhold=0.5 : F1=0.09 ; Recall=63.29% ; Precision=4.78%

同问，请问解决了吗？

[anova-viana]

0.09% = A类AP ||score_threhold=0.5 ： F1=0.00;回忆率=60.78%;精度=0.10% 3.10% = B类AP ||score_threhold=0.5 ： F1=0.08;回忆率=65.17%;精度=4.33% 3.33% = ClassC AP ||score_threhold=0.5 ： F1=0.09;回忆率=63.29%;精度=4.78%

两年了，同问