cv2.resize blocks forever when called by multiple processes(with multiprocessing module) #69
Description
Expected behaviour
N/A
Actual behaviour
I encountered a strange behavior of cv2.resize. I prepared a snippet that reproduces the strange behavior. pool.map blocks forever in the code below because any of those processes doesn't finish its job. The reason why I see it strange is that the code works okay when pool.map is not followed by "Single processing" code.
Steps to reproduce
#example code
import cv2
import multiprocessing
def process_image(image_path):
proc = multiprocessing.current_process().name
print proc, "processing image.."
img_input = cv2.imread(image_path, 0)
scale = 1.2
cv2.resize(img_input, (0, 0), fx=scale, fy=scale)
print proc, "finished"
def main():
image_path = '.../image.jpg'
# Single processing. If you comment out the line bellow, main function works just okay
process_image(image_path)
# Multi processing
num_processes = 5
pool = multiprocessing.Pool(num_processes)
pool.map(process_image, [image_path]*num_processes)
if __name__ == '__main__':
main()
-
operating system / architecture (e.g. x86)
Ubuntu 16.04 / x86_64 -
version
python 2.7.14(Anconda)
opencv-python==3.2.0.6