Upload an image of a document to extract its heading hierarchy.
Note: This is a demo version using an untrained model. The actual model would need to be trained on DocLayNet data for accurate results.