For scanned documents, especially old ones, some table lines can be broken, and that can be very hard for typical table OCR libraries to read! in this repository I deal with this issue, by first detecting the table using layout parser library, then preform OCR on the whole document , detect the contours and cells of the table, and finally assign the texts using their coordinates to their respective cells. I found that this method works much better then applying OCR to each detected contour cell