I am trying to get the numbers of the sports vehicles from their photos taken while they were in action.
My approach is: first detect the number plates using Yolo, cropping that part from the original photo(snapshots attached), and then OCRing on it to get the numbers.
The problem is with the ocr part as you can see the images are not ideal for any ocr (i.e. contain noise, different fonts, skewed etc.). I have tried preprocessing it with many different options available in openCV, also I have enhanced these using various DNNs and I have also used custom ocr. Nothing helped and I am still stuck with this issue.
Any suggestion/solutions would be helpful.