Ask Your Question

Can someone explain me these terms (acceptanceRatio, HR, FA) in training haar cascade, and "acceptanceRatio 800:4.97153e-05" is it consider good?

asked 2016-12-29 02:21:31 -0500

jugalRupela gravatar image

image description

edit retag flag offensive close merge delete

3 answers

Sort by ยป oldest newest most voted

answered 2019-07-25 02:41:32 -0500

Fadwa gravatar image

May be it is a bit late to answer this question but I was searching for answers to it and found this

Shortly: HR and FA stand for hit rate and false alarm. Conceptually: hitRate = % of positive samples that are classified correctly as such. falseAlarm = % of negative samples incorrectly classified as positive.

edit flag offensive delete link more

answered 2020-03-29 06:10:32 -0500

ComputerVisionary gravatar image

Take a look here on how to handle the acceptanceRatio and the False acceptanceRatio in opencv traincascade:

More about cascades:

edit flag offensive delete link more

answered 2018-03-06 04:01:54 -0500

Joshitha gravatar image

The precision of your cascade classifier is determined by your Acceptance Ratio of the last stage. Technically, the acceptance ratio break value tells how much your model should continue to learn and when to stop.It must ideally be around 0.0000412 or so.

If it is 4.8789e-05, it signifies that your cascade is overtrained and will not detect the objects. In this case, you will have to reduce the number of stages you set and increase the number of learning samples( give in more negative and positive images)

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2016-12-29 02:21:31 -0500

Seen: 2,075 times

Last updated: Mar 29 '20