Understanding document categorization with computer vision
In document preprocessing, you will come across use cases where documents are classified based on a branded logo on the document or having tables. These are visual clues that we want to use to classify documents before processing. We can use Amazon Rekognition, computer vision software with deep learning-powered image recognition, to detect visual clues such as objects, scenes, and text from any scanned images for document classification.
You can leverage an Amazon Rekognition Custom Label to detect a logo from any document and classify the document based on the logo. For example, a healthcare provider supports multiple insurance providers. Patients when visiting a doctor’s office submit insurance cards. These insurance cards can be processed automatically to detect the logos from them, and the documents can be classified according to their corresponding categories.
Now, let’s look at a hands-on example...