daxcay / comfyui-dataset Goto Github PK

Getting this error on big datasets

Request for workflow to automate the matching of SD training Captions in diffirent folders

The Workflow scans a parent folder repositary of the complete dataset of images and captions with same filenames.
It then scans subfolder with the main path, and analyses the filenames of the subset of images.
Lastly, the workflow will then copy over the matching caption files from parent folder to the subset folder.

As a result the subset folder will contain all the caption txt files that match the names of images in subfolder.

This is extremely useful in training with edited datasets. Sometimes datasets have hundreds, or thousands of paired image-captions. This workflow removes the need to find captions one by one with a human worker. DRMN together with JCDN are perfect tools to succeed at speeding up preparing datasets