OpenCV color_bgr2gray 错误

Question

There are a bunch of questions about this function throwing errors, such as here and here but I think mine has a unique twist. What I'm trying to do is go through all the images in a folder and detect any faces.

import cv2
import sys
import os

#Get the folder to be searched and the location of the cascade file from arguments
imageDirectory = sys.argv[1]
cascadePath = sys.argv[2]
faceCascade = cv2.CascadeClassifier(cascadePath)

#Iterate over the png files in the directory
for imagePath in os.listdir(imageDirectory):
    if imagePath.endswith(".png"):
        image = cv2.imread(imagePath)
        gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

        faces = faceCascade.detectMultiScale(
            gray,
            scaleFactor=1.1,
            minNeighbors=5,
            minSize=(30, 30)
            #flags = cv2.CV_HAAR_SCALE_IMAGE
        )

The first image will be processed correctly, but then the second will throw an error:

OpenCV Error: Assertion failed (scn == 3 || scn == 4) in cvtColor, file /tmp/opencv3-20170202-24427-1s95vpr/modules/imgproc/src/color.cpp, line 9748 Traceback (most recent call last): File "face_detect_flip.py", line 22, in gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) cv2.error: /tmp/opencv3-20170202-24427-1s95vpr/modules/imgproc/src/color.cpp:9748: error: (-215) scn == 3 || scn == 4 in function cvtColor

The files are all color .pngs so I can't tell why the script will work on some of them but not others. Sometimes changing the order or the filename (but not the extension) will change whether the script runs or not, but not in any pattern that I can discern.

Answer 1

Looks like the image that causes this error does not have 3 (rgb image) or 4 (rgba image) channels. The cvtColor function needs the the image to have at least 3 channels to convert it.

One possible reason could be that the file isn't being read correctly. This would cause the variable image to be None and hence it wont have the correct number of channels.

Another reason could be if the image has fewer channels. Maybe it's a grayscale image to begin with and thus has only one channel.

You can use print(image.shape) to check the number of channels. If the value is None then the file wasn't read correctly.