Revision history [back]

I’m having some issues with USB video capture devices and OpenCV and am looking for a solution.

I am using OpenCV v3.4.2 in a Python 3.7 script on two different Windows 10 platforms.

• Laptop that has a 2.8GHz Intel Core i7 processor and Intel HD 520 graphics adapter
• Desktop computer with 3.5GHz Intel Core i7 and Nvidia GeForce TX 660 graphics adapter

The video signal is an NTSC standard definition stream coming from a security camera. I have three different capture cards:

• Hauppauge USB Live-2
• A generic HD capture device with HDMI in. This device is attached to an SD-to-HDMI up converter (signal > converter > HD capture)
• A generic cheap standard definition capture device

My Python script instructs OpenCV to use the DirectShow API, but I believe OpenCV would default to DirectShow anyway when on Windows 10. Here are the results I’m getting for each of the three capture devices:

• The Hauppauge gives similar results on both computers. It often takes 3 or 4 restarts of the script before I get a good display. The first few attempts are often a black image in the video window.
• The HD capture device connects quickly with a good video on the laptop (it is horizontally stretched to fill out a 16:9 aspect ratio, but that is fine for my application). However, the video flickers a lot (unacceptable amount) on the desktop computer.
• The generic cheap SD capture device doesn’t work on either computer. I’m not sure it has a DirectShow API.

Here is a sample script.

import cv2
cap = cv2.VideoCapture()
# The device number might be 0 or 1 depending on the device and the webcam
cap.open(0, cv2.CAP_DSHOW)
while(True):
    ret, frame = cap.read()
    cv2.imshow('frame', frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
cap.release()
cv2.destroyAllWindows()

I want to use OpenCV in a Python script to reliably display a live video stream. Questions:

1. Is there a way to reliably get a good video display the first time when using the Hauppauge USB Live-2 capture device?
2. Why does the HDMI capture device and upconverter work on one computer but not the other when using the same Python script?
3. Is there a known solution for how to get a good video display on both computers when using the HD capture device with the upconverter?
4. Is there a recommendation for a reliable capture device? Standard definition NTSC is my first priority and HD (720p and/or 1080p) is second priority. I’m not concerned about 4K.

I’m having some issues with USB video capture devices and OpenCV and am looking for a solution.

I am using OpenCV v3.4.2 in a Python 3.7 script on two different Windows 10 platforms.

• Laptop that has a 2.8GHz Intel Core i7 processor and Intel HD 520 graphics adapter
• Desktop computer with 3.5GHz Intel Core i7 and Nvidia GeForce TX 660 graphics adapter

The video signal is an NTSC standard definition stream coming from a security camera. I have three different capture cards:

• Hauppauge USB Live-2
• A generic HD capture device with HDMI in. This device is attached to an SD-to-HDMI up converter (signal > converter > HD capture)
• A generic cheap standard definition capture device

My Python script instructs OpenCV to use the DirectShow API, but I believe OpenCV would default to DirectShow anyway when on Windows 10. Here are the results I’m getting for each of the three capture devices:

• The Hauppauge gives similar results on both computers. It often takes 3 or 4 restarts of the script before I get a good display. The first few attempts are often a black image in the video window.
• The HD capture device connects quickly with a good video on the laptop (it is horizontally stretched to fill out a 16:9 aspect ratio, but that is fine for my application). However, the video flickers a lot (unacceptable amount) on the desktop computer.
• The generic cheap SD capture device doesn’t work on either computer. I’m not sure it has a DirectShow API.

Here is a sample script.

import cv2
cap = cv2.VideoCapture()
# The device number might be 0 or 1 depending on the device and the webcam
cap.open(0, cv2.CAP_DSHOW)
while(True):
    ret, frame = cap.read()
    cv2.imshow('frame', frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
cap.release()
cv2.destroyAllWindows()

Note: The Python script works fine when the video source is the laptop's internal webcam.

I want to use OpenCV in a Python script to reliably display a live video stream. Questions:

1. Is there a way to reliably get a good video display the first time when using the Hauppauge USB Live-2 capture device?
2. Why does the HDMI capture device and upconverter work on one computer but not the other when using the same Python script?
3. Is there a known solution for how to get a good video display on both computers when using the HD capture device with the upconverter?
4. Is there a recommendation for a reliable capture device? Standard definition NTSC is my first priority and HD (720p and/or 1080p) is second priority. I’m not concerned about 4K.