目录
STEP_1:文件夹材料准备
# 文件
root/
neg/**.bmp 负样本
pos/**.bmp 正样本
opencv_createsamples 正样本转换
opencv_traincascade 正负样本训练
STEP_2:数据转换
# 数据设置{尺寸、名字、写入文件txt、目标ROI位置替换}
## 批量【 改变大小 】
```
for x in *.jpg; do convert -resize 256x256\! $x $x; done
```
## 批量【 改变名字 】
```
rename "s/^/image_/" x.bmp
```
## 批量【 改变名字 】
```
i=0;for x in *; do mv $x $i.bmp; let i=i+1; done
```
## 批量【 写名字名到文件 】
```
ls ./neg/*.* >neg.txt
ls ./pos/*.* >pos.txt
```
## 修改 pos.txt >>> xx.bmp 1 x y w h <<<
## 完成样本到文件
STEP_3:数据生成vec文件
-num 7092 正样本数量的大小
-w 40 -h 60 按目标的大小进行规划
## 修改样本截取大小
sudo ./opencv_createsamples \
-info pos.txt \
-bg neg.txt \
-num 7092 \
-vec pos.vec \
-w 40 -h 60
STEP_4:数据训练
手动建文件夹model,-data model
-w 40 -h 60 必须和上面规划的尺寸一致
-featureType LBP 有两种模型可选:LBP 和 HAAR
## 创建 model 文件夹、样本数量、尺寸、LBP 或者 HAAR、占用内存、准确率、错误率、使用的模型
sudo ./opencv_traincascade \
-data model \
-vec pos.vec -bg neg.txt \
-numPos 1000 -numNeg 1000 \
-numStages 20 \
-w 40 -h 60 \
-featureType LBP \
-precalcValBufSize 12400 -precalcIdxBufSize 12400 \
-minHitRate 0.9999 -maxFalseAlarmRate 0.25 \
-mode ALL
STEP_5:模型读取检测
python objectdetection.py 打开相机进行检测
python objectdetection.py --image = xx.jpg 读取图片进行检测
python objectdetection.py --video = xx.mp4 读取视频进行检测
# -*- coding: utf-8 -*-
import cv2
import argparse
import sys
import numpy as np
parser = argparse.ArgumentParser(description='Object Detection')
parser.add_argument('--image', help='Path to image file.')
parser.add_argument('--video', help='Path to video file.')
args = parser.parse_args()
# Process inputs
winName = 'Object detection in OpenCV-DJH'
cv2.namedWindow(winName, cv2.WINDOW_NORMAL)
outputFile = "out.avi"
if (args.image):
# Open the image file
if not os.path.isfile(args.image):
print("Input image file ", args.image, " doesn't exist")
sys.exit(1)
cap = cv2.VideoCapture(args.image)
outputFile = args.image[:-4]+'out.jpg'
elif (args.video):
# Open the video file
if not os.path.isfile(args.video):
print("Input video file ", args.video, " doesn't exist")
sys.exit(1)
cap = cv2.VideoCapture(args.video)
outputFile = args.video[:-4]+'out.avi'
else:
# Webcam input
cap = cv2.VideoCapture(0)
# Get the video writer initialized to save the output video
if (not args.image):
vid_writer = cv2.VideoWriter(outputFile, \
cv2.VideoWriter_fourcc('M','J','P','G'), 30, \
(round(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),\
round(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))))
if __name__ == '__main__':
## >>>>> 其他 <<<<<< ##
Cascade = cv2.CascadeClassifier(r'./model/cascade.xml')
while cv2.waitKey(1) < 0:
# get frame from the video
hasFrame, frame = cap.read()
# Stop the program if reached end of video
if not hasFrame:
print("Done processing !!!")
cv2.waitKey(3000)
cap.release()
break
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
objs = Cascade.detectMultiScale(gray,
scaleFactor=1.15,
minNeighbors=5,
minSize=(5,5),
flags = cv2.CASCADE_SCALE_IMAGE)
for (x,y,w,h) in objs:
frame = cv2.rectangle(frame, (x,y), (x+w,y+h), (255,0,0),2)
# Write the frame with the detection boxes
if (args.image):
cv2.imwrite(outputFile, frame.astype(np.uint8))
else:
vid_writer.write(frame.astype(np.uint8))
cv2.imshow(winName, frame)
结果展示
检测效果还不错,有车的模型、人脸的模型
STEP_6:训练经验
1. w h 按实际物体比例
2. storage: 看样本数量,多则多,少则少
3. acceptanceRatio: 在0.0004(4×10^3)左右寻找最佳的模型