-1

I have images that aren't timestamped and I need to rename and timestamp them for a project.

Here's an example of the timestamp, written at the top of every image:

enter image description here

So the question is, is there a way I can translate (detect) each number on the timestamp in the image shown? Perhaps I could use cv2 or tensorflow to do this?

Also I was thinking even training a decision tree if I were to crop each digit and create a series of same sized 1 channel arrays to train on.

Thoughts?

Miki
  • 40,887
  • 13
  • 123
  • 202
DavimusPrime
  • 368
  • 4
  • 17

2 Answers2

2

Why don't you use a simple OCR algorithm? The numbers and letters in the image is very clear, which I think an OCR algorithm will work fine.

A simple test with https://ocr.space/ produces the following:

****** Result for Image/Page 1 ******
14042018 Ph.n

For sure, training a model will be helpful if you have the corresponding labels.

lenhhoxung
  • 2,530
  • 2
  • 30
  • 61
1

When you only have computer-written text, you might want to test pytesseract for OCR.

Sebastian Rehfeldt
  • 41
  • 4