Development of Dynamic Android Library for Validation of OCR Data and Structured Data

Project Summary
Develop an android library to compare data extracted from a document image via OCR with data passed via an additional JSON structure. A simple comparison of the OCR extracted data and the data in the additional JSON structure will allow for a PASS/FAIL metric to be assessed, establishing how similar the two data sets are.

Document images that will have OCR applied will typically captured in the field and sourced from an on phone camera then passed to the library for processing. These images may have inconsistent quality and resolution, however the specific image area to be processed by OCR is well defined and there is flexibility in font selection, text point size and text positioning. We expect a developer experienced with OCR to offer suggestions on text/OCR specifics and possible options to maximize OCR success and accuracy rates. All processing must be accomplished on-device.

This library will be included in a larger Android offering and represents a core service of the larger project.

Key skills required – OCR experience, Android library development.

Scope and Deliverables
• Developer should recommend implementation source language.
• Developer should suggest appropriate dependency libraries.
• Developer should suggest/define library API call format.
◦ API should be simple but flexible/extensible for future enhancement to the library.
• Developer must produce a fully documented standalone (statically linked dependencies) dynamic library.
• Developer must produce an additional functional example program which shows correct library usage in a test case scenario.
• Developer must deliver a prebuilt set of library binaries (multi-arch) and all source code, including source code for build dependencies.
• Developer must release all applicable copyright assigns, full source code and complete working/functional build chain on completion of this commission.
• Developer is required to use Libre/Open/Free dependencies (GPL/LGPL/MIT or more permissive)