OpenCV set() and read() function is too slow (Python)

Question

I want to loop frames from a certain range, and append it to an array. The thing is that, it's too slow. I already check how long the function would take and i think that's pretty slow. Here is my current code:

imgs = []
for j in range(range1, range2):
      video.set(cv.CAP_PROP_POS_FRAMES, j)
      ret, frame = video.read()
      imgs.append(frame)

i also tried to replace imgs.append(frame) with video.retrieve(video.grab()), but the performance didn't really differs much. Is there any better solution/alternative to do what this code does??

You could try to enable CUDA to use GPU to speed up. Not sure if it will actually do alot of speedup but learnopencv.com/getting-started-opencv-cuda-modul — Jason Chia
– Jason Chia, Commented Jan 22, 2021 at 14:56
I don't think the set is the slow part, but you can get rid of it by moving it outside of the for loop. You can set it to range1 before the for loop, video.read() will advance by one frame each time it's caled. — Ian Chu
– Ian Chu, Commented Jan 22, 2021 at 15:21

Ian Chu · Accepted Answer · 2021-01-22 15:31:05Z

3

Oh wow, nvm. It is set that's making this slow:

Inside Time: 15.308052062988281

Outside Time: 0.4459998607635498

import cv2
import time

def setInside(cap, start, end):
    imgs = [];
    for a in range(start, end):
        cap.set(cv2.CAP_PROP_POS_FRAMES, a);
        _, frame = cap.read();
        imgs.append(frame);

def setOutside(cap, start, end):
    imgs = [];
    cap.set(cv2.CAP_PROP_POS_FRAMES, start);
    for a in range(start, end):
        _, frame = cap.read();
        imgs.append(frame);

# open vidcap
cap = cv2.VideoCapture("202534.avi");

# bounds
start = 0;
end = 2000;

# time it
start_time = time.time();
setInside(cap, start, end);
print("Inside Time: " + str(time.time() - start_time));

start_time = time.time();
setOutside(cap, start, end);
print("Outside Time: " + str(time.time() - start_time));

If you move the set to before the loop it'll be way faster.

answered Jan 22, 2021 at 15:31

Ian Chu

3,14311 silver badges16 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

dwiandhika Over a year ago

thank you very much, it works!. For me it reduces the time required from 3-3.5 sec to 0.5-0.8 sec.

dwiandhika Over a year ago

Btw, is there any way to squeeze the performance even more by replacing the cap.read() function with something else, maybe cap.retrieve() or something? just asking..

Ian Chu Over a year ago

I don't think so. cap.read() is really just a wrapper that combines cap.grab() and cap.retrieve() into one function.

Chuck Liao · Accepted Answer · 2021-07-03 03:35:35Z

0

I also came across this problem. I have to use set(3,width) and set(4,height)to acquire the original size of my camera but this method makes my code run much slower

answered Jul 3, 2021 at 3:35

Chuck Liao

1

1 Comment

tomasantunes Over a year ago

This is not an answer. See How do I write a good answer?

Collectives™ on Stack Overflow

OpenCV set() and read() function is too slow (Python)

2 Answers 2

3 Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related