Introduction
Our team Entanglo(Geunyoung Gil, Marcus Rankin) created web application to collect data automatically from images. We came to this idea to make people easy to collect their data with less efforts. It is single webpage application, so people can use their mobile to collect any text data from their images. Our team was awared 1st place on the final capstone projects competition of Software Engineer Technologies of Conestoga College at Apr.28.2018. This compition was broadcasted by CTV news. video link.
Source Code
Demonstration Video
Please, click video link to see video how to use.
Process
I will explain one by one how it works.
Our application is all touch based. Slide to move from entry page.
Take a pictue to grab text and classify the words to entities.
It captured data through Google OCR and classified entities with Watson and our own algorithm.
User can select items which user want to collect. But if you look at some of selected items,
it didn’t detect entities. User can manually combine items and type the entity like below.
This image show the result of manually combined and typed items. (You can check in video link in detail.)
Discard items which user doesn’t want to collect and save items to database.
Check items from database.
After this, we assume that user wants to save another business card. User take a picture of another business card,then our app will automatically collect data which user selected items only before(we did above). You can check on video link.
Application design
Our solution is composed of 2 web api applications which are “Entanglo Service API” and “Entanglo Website”.
[Sevice API application] It provides APIs (Asp.net core) that do CRUD for data collection with PostgresSQL for each user. It is stand-alone service. So it can be used for different web application. We initially designed to use for various web applications for future.
[Entanglo Website] It has Angular front-end with Three.js for UI and Asp.net core back-end. The back-end service use the apis from Entanglo Service, Google OCR, and Watson entity classification to recognize text from images and classify entities from text.