CRAFT training on non ocr images

[gupta9ankit5]

Hi,

This is regarding training the CRAT model (the detection segment of EasyOCR). Apart from images containing text as part of the dataset, I also have images with no text, and I want the model to be trained on both types. While label files are provided for images containing text, I am unsure how to create labels for images without text.

Do I need to create label files for such images, or should the label file be created but left blank?