Skip to content
This repository has been archived by the owner on Sep 3, 2024. It is now read-only.

ffmpeg error when create_dataset.py #66

Open
MagicLizi opened this issue Jan 9, 2020 · 3 comments
Open

ffmpeg error when create_dataset.py #66

MagicLizi opened this issue Jan 9, 2020 · 3 comments
Assignees
Labels
bug Something isn't working

Comments

@MagicLizi
Copy link

  • OS: macOS
  • Device CPU
  • Setup Info PyTorch 0.4.1, Python 2.7

When I deal create_dataset.py
ffmpeg report so much erorr as
Application provided invalid, non monotonically increasing dts to muxer in stream 0: 1 >= 1

image

@MagicLizi MagicLizi added the bug Something isn't working label Jan 9, 2020
@xiedidan
Copy link

xiedidan commented Feb 6, 2020

Setting ffmpeg -vsync to 1, 2, or drop seems to partly solve this problem. But they all drop frames if any timestamp (DTS/PTS) is incorrect.
With -vsync drop, I got 6947 training clips from adobe240fps.
Even with the latest ffmpeg master branch, this problem persists. I suspect we couldn't reproduce the same result as the author's. I'll try to modify dataset creation script to extract frames with opencv video stream...

@xiedidan
Copy link

xiedidan commented Feb 7, 2020

Opencv extracted 8488 training clips, there are two videos dropped most of their frames - IMG_0179.m4v and GOPR9658.MP4, while most of the videos drop a small partion of frames.
Since IMG_0179.m4v and GOPR9658.MP4 do look problematic with MPC, I'm happy with the opencv results.

@devanshbatra04
Copy link

@xiedidan can you please share what modifications are to be done to use OpenCV? I am running into the same issue.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

4 participants