OpenCV无法获得haarcascade

Question

I'm writing a Python Script using OpenCV library. The code works flawlessly, except one bit. I'm going to build the script with pyinstaller so I need to reference the haarcascade.

face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_alt.xml')

This works, but this:

def resource_path(relative_path):
    try:
        base_path = sys._MEIPASS
    except Exception:
        base_path = os.path.abspath(".")

    return os.path.join(base_path, relative_path)


face_cascade = cv2.CascadeClassifier(resource_path('haarcascade_frontalface_alt.xml'))

doesn't. Need help with solving this problem

Here is the problem: the haarcascade won't load and the "detectMultiScale" will fail.

The same solution worked for my another project

Here's the eror I'm getting: I tried using "face_cascade.load()" And it works like this:

face_cascade.load('haarcascade_frontalface_alt.xml')

But again doesn't work with "resource_path" function

Doing this

import os.path
print(os.path.isfile(resource_path('haarcascade_frontalface_alt.xml')))

Prints "True" in the console

Also I tried deleting all the code related to detecting faces, and I found that OpenCV successfully captures my webcam

Here is all the code: if someone's interested

import numpy as np
import cv2, os



cap = cv2.VideoCapture(0)
xe = 0
ye = 0
we = 0
he = 0
def resource_path(relative_path):
    try:
        base_path = sys._MEIPASS
    except Exception:
        base_path = os.path.abspath(".")

    return os.path.join(base_path, relative_path)


import os.path
print(os.path.isfile(resource_path('haarcascade_frontalface_alt.xml')))



face_cascade = cv2.CascadeClassifier(resource_path('haarcascade_frontalface_alt.xml'))
face_cascade.load(resource_path('haarcascade_frontalface_alt.xml'))


while 1:
    ret = cap.set(3,640);
    ret = cap.set(4,480);
    ret, img = cap.read()

    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray, 1.2, 5)
    dst = img

    for (x,y,w,h) in faces:
        dst = img
        rows,cols,channels = img.shape
        xe = x
        ye = y
        we = w
        he = h

    rows,cols,channels = img.shape
    pts1 = np.float32([[xe-100,ye-100],[xe+2*we+200,ye-100],[xe-100,ye+2*he+200],[xe+2*we+200,ye+2*he+200]])
    pts2 = np.float32([[0,0],[cols,0],[0,rows],[cols,rows]])
    M = cv2.getPerspectiveTransform(pts1,pts2)
    dst = cv2.warpPerspective(img,M,(300,300))
    dst = cv2.resize(dst,(cols, rows), interpolation = cv2.INTER_CUBIC)

    cv2.imshow('img',img)
    k = cv2.waitKey(30) & 0xff
    if k == 27:
        break

cap.release()
cv2.destroyAllWindows()

Edit:

I found the reason for the error: the pass to haarcascade contained cyrillic symbols that's why OpenCV had difficulties with loading the file

Answer 1

Is the haarcascade file actually in the place you expect it to be?

In your resource_path function you could check whether the file actually exists at the full path that is created to be sure.

Answer 2

Get rid of all cyrillic symbols in the path to haarcascade if you get this problem

Answer 3

Okay so even I was facing the same problem. the thing is you have to externally provide the directory link of haarcascade to the classifier function.

for me using anaconda it was at this directory:

C:\Users\lhari\anaconda3\pkgs\libopencv-3.4.2-h20b85fd_0\Library\etc\haarcascades

Then you can join your paths and put it like below as it worked for me!!

path = os.path.join(r"<Your directory to the opencv package>",  r"haarcascade_frontalface_default.xml")
face_classifier = cv2.CascadeClassifier(path)

Answer 4

What I can understand is that you need to pass the facecascade file while building the exe.

TRY THIS:

pyinstaller <filename> --add-data="<location of cascade>:.>"

by running this the application will have the cascade file. more details can be found here