TrainCascade image positive problem
I've create a classifier using Haartraining_Stuff, i have to create a road signal detection, i create a file info, negative.txt, positive.txt and .vec file. Then i use command traincascade write like this:
C:\OpenCV\opencv\build\x86\vc11\bin\opencv_traincascade.exe -data classifier\ -vec "C:\OpenCV\sample.vec" -bg "C:\OpenCV\negativeNew.txt" -numStages 2 -minHitRate 0.01 -maxFalseAllarmRate 0.5 -numPos 2 -numNeg 250 -w 20 -h 20 -mode ALL -precalcvalBuffSize 256 -precalcdxBufSize 256
But when i run it, give me this:
C:\OpenCV>C:\OpenCV\opencv\build\x86\vc11\bin\opencv_traincascade.exe -data classifier\ -vec "C:\OpenCV\sample.vec" -bg "C:\OpenCV\negativeNew.txt" -numStages 2 -minHitRate 0.01 -maxFalseAllarmRate 0.5 -numPos 2 -numNeg 250 -w 20 -h 20 -mode ALL -precalcvalBuffSize 256 -precalcdxBufSize 256 PARAMETERS: cascadeDirName: classifier\ vecFileName: C:\OpenCV\sample.vec bgFileName: C:\OpenCV\negativeNew.txt numPos: 2 numNeg: 250 numStages: 2 precalcValBufSize[Mb] : 1024 precalcIdxBufSize[Mb] : 1024 acceptanceRatioBreakValue : -1 stageType: BOOST featureType: HAAR sampleWidth: 20 sampleHeight: 20 boostType: GAB minHitRate: 0.01 maxFalseAlarmRate: 0.5 weightTrimRate: 0.95 maxDepth: 1 maxWeakCount: 100 mode: ALL
===== TRAINING 0-stage ===== < BEGIN POS count : consumed 2 : 2 NEG count : acceptanceRatio 250 : 1 Precalculation time: 0.594 +----+---------+---------+
| N | HR | FA |
+----+---------+---------+
| 1| 1| 0|
+----+---------+---------+ END> Training until now has taken 0 days 0 hours 0 minutes 1 seconds.
===== TRAINING 1-stage ===== BEGIN POS count : consumed 2 : 2 NEG count : acceptanceRatio 250 : 0.0423155
Required leaf false alarm rate achieved. Branch training terminated.
I think the problem is the low positive image no?
There is a command for use distorsion or something like this for create more file positive without take photo around? Thanks for help.
Are you sure for:
Also there is a typo error:
-maxFalseAllarmRate
.Both values seem pretty low for me but people with better understanding on Cascade Classifier would be able to answer you.