How to read a video file with opencv

[AlejandroDiazD]

Hello @Corallo ,

I'm still trying to solve the integration with Yolov5. meanwhile, I'm trying some tests with video files, but I'm not being able to open a video from a python script with opencv. I can open an image, but I'm trying to open a video following the standard opencv steps:

import cv2
video_path = "video_path.mp4"
video = cv2.VideoCapture(video_path)
video.read()

I've tried with bot codec formats, H264 and MJPEG, but always obtaining the same results:

(False, None)

Is there any available coded on the sdk or any other way to open a video file?

Thank you!

[AlejandroDiazD]

Extending the previous question, I would have the same doubt about if is there any available codec for saving a video file from streaming using a python script, as I only saw in the object-detector-python example that images are saved but not any video.

Thank you again!

[Corallo]

Hi @AlejandroDiazD

We don't have an example of how to parse a video with opencv.
It is reasonable to think that as you said it is a problem of some codec missing.
Your application container is just an Ubuntu image, if you need more codec, you can add an RUN apt install ... in the docker image to install what you need.

[AlejandroDiazD]

Ok, I will try that. Thank you!

Btw, I'm also trying to use the camera streaming as input and again with wrong results...

I followed the steps of the object-detector-python example for reading input streaming video from the camera but the image I'm saving is so blurred to be used (and also seems to be horizontally duplicated 4 times in the case of setting 4:3 format from the camera interface, and 16 times in the case of 16:9).

This an example of an image received, setting 4:3 format with 1280x960 resolution:

If I see the image on the web interface is perfect:

I have also tried reading the image streaming with the steps:

import cv2
cap = cv2.VideoCapture(0)
cap.read()
Explained in https://axiscommunications.github.io/acap-documentation/docs/api/computer-vision-sdk-apis.html#video-capture-api

But again the results are:

(False, None)

Thank you for your help!

[Corallo]

The second methods is deprecated (using opencv), the documentation update had a delay, today or tomorrow is going to be updated.
But the other thing you showed seems to be an issue.
Could you open an issue on the repository with a minimal reproducible code example? Also, please mention the resolution, camera model and firmware you are using. We'll look into it

[AlejandroDiazD]

Done! #181

Thanks!