2.0 KiB
2.0 KiB
PairedImageLoader
PairedImageLoader pulls matching source/processed image pairs from aligned folders. It’s ideal for before/after reviews or comparison pipelines where both streams must advance together.
Inputs
source_folder– Directory containing the “before” images.target_folder– Directory containing the “after” images.pattern– Optional filename template (e.g.,{name}_after.png) to align variations when the names differ.auto_next– Step to the next pair on every execution.loop– Restart from the top when you reach the end of the folder.shuffle– Randomise the order of the pairs.strict_matching– WhenTrue, only emits pairs where both files exist. Disable to allow singletons.
Outputs
source_image– The “before” image tensor.target_image– The matching “after” image tensor.source_filename,target_filename– Filenames currently loaded.index– Position in the list for logging or syncing with other nodes.
Where It Fits
Use PairedImageLoader for A/B pipeline QA, training-data spot checks, or client review decks where each processed shot should sit next to its original capture.
Tuning Tips
- Keep
strict_matchingenabled during quality checks so missing files are obvious. - Combine
auto_nextwithComparisonGateand your preferred viewer for smooth slideshow-style reviews. - Use the
patternfield to translate naming schemes, such as turningportrait.jpgintoportrait_graded.png.
Troubleshooting
- Pairs go out of sync – Double-check folder naming and the
patternformat. The node advances both sides together when matches are found. - Only one image appears – If strict matching is on, missing companions stop the pair. Either add the file or disable strict mode to let singletons through.
- Order unexpected – Toggle
shuffleoff and ensure the folders list files in the same sequence.
Screenshot: docs/screenshots/portraitutils_paired_image_loader.png