Closed ducviet00 closed 3 years ago
increase the score threshold, filter out the low confidence scores ... the score threshold for eval metrics is really really low. It is not suitable for visualization.
increase the score threshold, filter out the low confidence scores ... the score threshold for eval metrics is really really low. It is not suitable for visualization.
Thanks
I've trained EfficientDet on my custom data. My testset has 586 images but when I run validate.py for evaluation, I get a lot of duplicate bbox - 58600 predicted bbox in result.json thanks