<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/06504f67-bb75-4f61-bb93-1771e33394a3/original=true/06504f67-bb75-4f61-bb93-1771e33394a3.jpeg" />Hi everyone! I’d like to share a batch processing workflow designed for LoRA training image datasets. It batch-reads images from a specified directory to apply upscaling, resizing, and captioning. The processed images and their corresponding caption text files will be renamed and saved to a directory of your choice. Best of all, you have the flexibility to pick and choose which of these three processes you actually want to run!How to Use:<ol><li>Input Path: Enter the input path for the images you want to process.</li><li>Trigger Word: Enter your trigger word (if needed). It will be prepended to the beginning of each caption after the captioning process is complete.</li><li>Output Path: Enter the output path for the processed results.</li><li>Execution Settings: Before executing, set the index to <code>0</code> and the batch count to the total number of images to be processed.</li></ol><hr />大家好！分享一个专为 LoRA 训练图像集打造的批处理工作流。它会批量读取指定目录下的图像，依次进行超分、Resize（尺寸调整）和 Caption（自动打标）处理。处理后的图像和对应的 caption 文本文件将被统一重命名，并保存到你指定的输出目录下。当然，你完全可以根据自己的需求，灵活选择要执行上述三个流程中的哪些步骤。使用说明：<ol><li>输入路径： 输入待处理图像所在的文件夹路径。</li><li>触发词（Trigger Word）： 输入你的触发词（如有需要）。在 Caption 阶段执行完毕后，该词会被自动追加到每段 caption 内容的最前面。</li><li>输出路径： 输入处理结果的保存路径。</li><li>执行前设置： 在点击执行前，请将 index（索引） 设置为 <code>0</code>，并将 batch count（批次数量） 设置为待处理图像的总数。</li></ol>

Files with SHA256: f336cb7cb2089c21853eaabc9c5ab79f66a040c40625f39d566bdeaa7671c407