I recently made a request to read images from Tesseract OCR. However, the problem is that the accuracy of reading numbers is only about 60%. Is there any way to improve api for reading numbers? Or should I stick with tesseract? I'm actually new to OCR and it's hard to handle.

source share