The keyframe image extracted is different from the keyframe image in the annotation software, and the number of frames is not synchronized

Python
1

Labeling software DarkLabel :point_right:https://github.com/darkpgmr/DarkLabel/blob/master/darklabel.yml

image
image1035×676 52.1 KB

The pictures captured by pthon are inconsistent with those of the annotation software, and the number of frames is different.
It feels like there are a few frame errors and the values ​​are not fixed. How should I adjust it? What’s the problem?

opencv-python 4.8.0.76

2

53f5f0b39529d24705405ac04243033
53f5f0b39529d24705405ac04243033379×626 67.5 KB

3

ef5cd6c288dcca95f6432b4be72c2f7
ef5cd6c288dcca95f6432b4be72c2f7379×626 60.7 KB

4

image
image836×196 10.4 KB

5

common issue with broken video files.

seeking is always guesswork to some extent.

transcode your videos to constant frame rate and intra-coded.

or maybe the annotation software is broken? submit a bug with that project instead.

as far as I can see, you haven’t demonstrated where the issue is. an [mre] consists of data. you need to provide data to reproduce the issue.

1 Like
6

The problem is that the code and the annotation software do not use the same encoder.The annotation software used is oencv_videoio_ffmpeg430.dll .How do I now get the code to handle the video to also use oencv_videoio_ffmpeg430.dll？Is there any official usage?

7

For example, the 199th frame is marked, but the 199th frame is intercepted in the code, which is actually the 196th frame. The difference is about 3 frames, and the error is not fixed.