OpenImagingLab / OpenImagingLab/FlashVSR

Support for longer videos?

Open
#27 1 comment 3 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
1.9k
Forks
152
PR merge metrics
No merged PRs in 30d

Description

Excellent work, but I noticed in the current project code that when processing videos longer than 10 seconds, the model runs out of memory due to accumulated features. However, I observed that the entire sequence is being saved at present. Would it be possible to set overlaps between segments—for example, 32 frames (to safely cover the temporal cache built by the model in the first segment)? From the code, it seems that prefetching within the first segment goes up to approximately frame 21, and subsequent blocks advance with +17/+21 indices. Each segment could be fed into the current script for inference independently, with saving and explicit cleanup performed at the end of each segment.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by tracing the current video-processing script, especially where the full sequence and accumulated features are saved. Test a video longer than 10 seconds and confirm the memory failure, then assess segmented inference with 32-frame overlaps and explicit cleanup; done means longer videos process without exhausting memory while preserving the expected output.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.