OCRestful Handwriting Recognition
OCRestful can perform handwriting recognition for you. Here's how it works:
Step 1. You start by uploading a sample document and using it to create a handwriting recognition template. On this template, you identify regions that contain handwritten fields to be captured, and optionally, you can identify individual fields that will be captured as discrete data elements later on.
Step 2. Next, your application uploads documents through the OCRestful REST API, using the handwriting recognition endpoint (ending with /icr/) and specifying the handwriting recognition template you previously created. (Note: you can optionally supply a webhook to be notified when your OCR / ICR results are ready).
Step 3. OCRestful will perform normal OCR on the main body of the document, but will treat the handwriting recognition regions specially. It will create an image of each region, and will initiate an Amazon Mechanical Turk HIT (Human Intelligence Task) to read and transcribe the data from the image regions into individual response fields.
Step 4. Amazon Mechanical Turk distributes the HITS to its human workers, who will transcribe the data from the images according to your template.
Step 5. Amazon returns nicely-formatted JSON data to OCRestful, which in turn makes it available for you to download to your application. If you set up a webhook in Step 2, it will be called when Amazon returns the data to OCRestful, which will be your application's notification that the handwriting data is ready to be collected.
Note: You control the structure of the JSON by how you set up your template.
A few additional notes about OCRestful handwriting recognition: