Commodity OCR Bakeoff

  In this challenge we'd like to explore the options around using "commodity" Optical Character Recognition (OCR). We have a large number of images we'd like to process at scale and we'd like to explore the viability of using a SaaS solution to do this. 

  For this challenge, almost any SaaS OCR system is fair game as long as it can be shown to process images to text outputs with good fidelity. IBM's Watson or Google TensorFlow are great places to start but are not required. Show us what you have. 

  Expected outputs are PoC code that shows how this could work as well as good documentation and description of the possible solution. 

  A link to some sample images that we'd like to target are in the forum. 

Final Submission Guidelines

Your submission should include any code you have written as well as documentation that describes how it should be deployed. In addition, you will be required to provide a one-two page powerpoint (or gdoc presentation) of the features and architecture of your solution. A quick video of your solution is highly desired but not mandatory. Your solution should deployed on a hosted service such as Heroku.