I am trying to implement HOG Descriptor with OpenCV to detect Pedestrians in a video. I am currently using the pre-made dataset by OpenCV hogcascade_pedestrians.xml. Unfortuntley the documentation on this part is very poor on the internet although the HOG Descriptor is very effective for human detection. I have been writing a code for pedestrians detection with Python, and I have stopped at the following code:
import cv2
import numpy as np
import imutils
VidCap = cv2.VideoCapture('pedestrians.mp4')
HOGCascade = cv2.HOGDescriptor('hogcascade_pedestrians.xml')
HOGCascade.setSVMDetector(cv2.HOGDescriptor_getDefaultPeopleDetector())
while True:
_ , image = VidCap.read()
image = imutils.resize(image, width=700)
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=15.0,tileGridSize=(8,8))
gray = clahe.apply(gray)
winStride = (8,8)
padding = (16,16)
scale = 1.05
meanshift = -1
(rects, weights) = HOGCascade.detectMultiScale(gray, winStride=winStride,
padding=padding,
scale=scale,
useMeanshiftGrouping=meanshift)
for (x, y, w, h) in rects:
cv2.rectangle(image, (x, y), (x+w, y+h), (0,200,255), 2)
cv2.imshow('Image', image)
if cv2.waitKey(5) == 27:
break
VidCap.release()
cv2.destroyAllWindows()
I presume that the code scripting would be something like codes written for Haar Cascades. But I have tried that and I got errors. Do anyone have any idea of how to implement the HOG Descriptor on OpenCV with Python.
I have read the following question, but I get nothing from the second answer.
My problem is that I can't find the way to write the code, as the documentation about this part is very poor.
Note: I am using OpenCV 3.1.0-dev with Python 2.7.11
解决方案HOGCascade = cv2.HOGDescriptor()
If you want to use this .xml, You have lots of preparation work to do.
When u finally get the available descriptor, you should replace the cv2.HOGDescriptor_getDefaultPeopleDetector() in
setSVMDetector(cv2.HOGDescriptor_getDefaultPeopleDetector())