Document Image Text Extraction With Layout Preservation

via “optical character recognition with layout preservation”

Microsoft's unified model for diverse vision tasks.

Unique: Performs end-to-end OCR with layout preservation using a single seq2seq model that generates text tokens interleaved with coordinate sequences, eliminating separate text detection and recognition stages

vs others: Simpler pipeline than Tesseract + text detection models but with 15-25% lower character accuracy on printed documents; stronger on handwriting and scene text than traditional OCR