Fix custom video decoding #769
Open
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix: Enable decoding of custom .mp4 videos in Kinetics-style datasets
Summary
This PR introduces support for decoding
.mp4videos in datasets following the Kinetics-style directory and CSV structure.Key Changes
.mp4files.decoder.pyfor torchvision compatibility.kinetics.py,loader.py, and__init__.pyto allow seamless loading of custom datasets without re-encoding.my_custom_dataset.pywhich can be reused by others for loading custom structured datasets.Usage
You can now train SlowFast models directly on
.mp4video files using a structure similar to: