OCR extraction from scans

Question

I really need your help!

I have been tasked by my university to digitise a whole bunch of historical data. These are old scans from books and my task would be to copy the data by hand.

Since this task annoys me a lot, I was wondering if I could use OCR in Python that would extract the data and create a csv file. In the meantime I have encountered the problem that

1. sometimes bad readability of the numbers
2. table lines are seen as characters: / ( ) By now I have tried Onenote character extraction and Excel 'read from image'. Both didn't go well.

So I'm wondering if anyone can recommend a package specifically for this data, or if you have tips and tricks for this task that would otherwise take me weeks.

One example of the scans:

Answer 1

Getting this to OCR direct to data will be likely next to impossible. The numbers are skewed, and the graphic components are inconsistent. So since this will take informed and willful processing, outside a human overlooking and cleaning, Maybe AI driven OCR https://cloud.google.com/use-cases/ocr could get you closer to "what you meant" vs "what was literal"

They have a free trial, may be worth a look.

Other than an AI driven solution, a temp typing, or maybe proofreading imperfect OCR scans. will probably be the only real solution.

Example this is the best I could get tesseract to do.

1 922 1 299

1 028 1 568

1 828

2 085 1771

1 205

2 147 1173

2 347 1 489 2 846

1 620 1 763 646

1579

693 519 466

317

816 817 2 364

1 182 1 054

869 2 958 1 201 1 049

816

81

151

18 10

0 21

j9Sbm.png (958x756)

1515

1730 1 152,

872. 141 1)

1 688 | 1 800! 1 448) H |

1 006) 764! 233 |

1 843| 1 130;

i 2 042) 1 264) 2 593)

1 396

561 1 37

623 566 41% He 34 mm

6. i 721

1 964) 1 05: 934