OCR Pipeline with YOLO, TrOCR and Roberta

Upload an image to detect text regions with YOLO, merge bounding boxes, and extract text using TrOCR which is then preprocessed with Roberta for contextual understanding.

image

Bounding Box Image

Extracted Text (Custom trained YOLO Object Detection + TrOCR Vision Transformer)

Post Processed Text (BLEU score based filtering + Roberta contextual understanding)