detection_eval is high but demo_det.py find a lot of false positive boxes

[MarStarck]

my detection_eval is about 0.77~0.86 according to different iterations. but the detection results contains around 50% false positive boxes.

[MhLiao]

You can enlarge the threshold score of the detection to filter false positive boxes.

[zhangxiaopang88]

many false positive boxes have scores above 0.9，Is the regression not good? can you give me any suggestions?

[mengzhangjian]

@zhangxiaopang88 , I also find that although detection_eval when trainging is very high, but when testing, I found the test result score is relatively low, do you know the reason?