I have built opencv from source detached at 3.4.6. And then I use /path_opencv_built/bin/opencv_traincascade to train a cascadeclassifier.xml.
opencv_createsamples -info pos.txt -num 10000 -w 100 -h 100 -vec pos.vec
opencv_traincascade -data data -vec pos.vec -bg neg.txt -numPos 9000 -numNeg 100 -w 24 -h 24 -numStages 15
I thought it should be right because I have got reasonable output.
PARAMETERS:
cascadeDirName: data
vecFileName: pos.vec
bgFileName: neg.txt
numPos: 9000
numNeg: 100
numStages: 15
precalcValBufSize[Mb] : 1024
precalcIdxBufSize[Mb] : 1024
acceptanceRatioBreakValue : -1
stageType: BOOST
featureType: HAAR
sampleWidth: 24
sampleHeight: 24
boostType: GAB
minHitRate: 0.995
maxFalseAlarmRate: 0.5
weightTrimRate: 0.95
maxDepth: 1
maxWeakCount: 100
mode: BASIC
Number of unique features given windowSize [24,24] : 162336
===== TRAINING 0-stage =====
<begin< p="">
POS count : consumed 9000 : 9000
NEG count : acceptanceRatio 100 : 1
Precalculation time: 2
+----+---------+---------+
| N | HR | FA |
+----+---------+---------+
| 1| 0.998556| 0.76|
+----+---------+---------+
| 2| 0.998556| 0.76|
+----+---------+---------+
| 3| 0.999444| 0.77|
+----+---------+---------+
| 4| 0.999889| 0.8|
+----+---------+---------+
| 5| 1| 0.8|
+----+---------+---------+
| 6| 1| 0.8|
+----+---------+---------+
| 7| 0.999| 0.41|
+----+---------+---------+
END>
Training until now has taken 0 days 0 hours 12 minutes 4 seconds.
...
...
===== TRAINING 14-stage =====
<begin< p="">
POS count : consumed 9000 : 9394
NEG count : acceptanceRatio 100 : 0.0045939
Precalculation time: 2
+----+---------+---------+
| N | HR | FA |
+----+---------+---------+
| 1| 0.998111| 0.76|
+----+---------+---------+
| 2| 0.998111| 0.76|
+----+---------+---------+
| 3| 0.999333| 0.76|
+----+---------+---------+
| 4| 0.996111| 0.44|
+----+---------+---------+
END>
Training until now has taken 0 days 1 hours 41 minutes 42 seconds.
And I have got files. (cascade.xml params.xml stage0.xml stage1.xml stage10.xml stage11.xml stage12.xml stage13.xml stage14.xml stage2.xml stage3.xml stage4.xml stage5.xml stage6.xml stage7.xml stage8.xml stage9.xml)
But when I use the cascade.xml, I always blocked using the method detectMultiScale(), without any responses even ctrl+c didn't work.
BTW, it runs well when I use origin haarcascade_fullbody.xml. I just don't know where the problem is.
My code shows below.
import cv2
import time
model_haar = cv2.CascadeClassifier("cascade.xml")
cam = cv2.VideoCapture('video.mp4')
print("load success.")
while True:
_, img = cam.read()
gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
print("read image")
t0 = time.time()
result = model_haar.detectMultiScale(gray_img)
print(time.time() - t0)
for resule_x, result_y, result_w, result_h in result:
cv2.rectangle(img, (resule_x, result_y), (resule_x + result_w, result_y + result_h), (255, 0, 255), 2)
cv2.imshow('img', img)
key = cv2.waitKey(30) & 0xff
if key == ord('q'):
break
cam.release()
cv2.destroyAllWindows()
Thanks for any advice~
