Beyond simple OCR: an outlook of modern AI vision for insurance automation

A typical OCR example
Is the checkbox ticked or not?
Is the document correctly signed? Does the signature correspond to the user database-registered signature?
Observe how characters seem to have been software edited
Were the two vehicles riding in the same direction? Did one of them cross a white line?
The new French driving license encapsulates fields with this weird symbol, known as the “Alien character”. While it is an effective way to prevent forgery, it will also confuse most OCR systems, as they will try to interpret it as a regular character. Here, ML algorithms have been used to remove them before going into OCR
An example of french redemption request in which the strokes that are printed on the form ruin the OCR prediction, typically interpreted as ‘I’ or handwritten characters continuation. Removing them with Computer Vision methods leads to an increased OCR accuracy
EfficientNet B3, one of the ConvNet featured in this blog post series

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store