Object Detection on Newspaper images using YoloV3

I was trying my hand on Optical Character Recognition on newspaper images when I realised that most documents have sections and text is not necessarily across the entire horizontal space of the page. Even though Tesseract was able to recognise the text it was jumbled up. To fix this the model should be able to identify sections on the document and draw a bounding box around it an perform OCR. It was this moment when applying Yolo Object detection on such images came into mind. YOLOv3 is extremely fast and accurate.