Images contain so much valuable data. But once an image is added into a system, it becomes a blackbox as it is not possible to what was in the image. OCR technology has been around for a while, so the idea was to bring it to life for Attachments for Confluence cloud. We have developed proprietary algorithms to make all the text to be extracted as traditional tools were only fetching dark text in consistent light backgrounds.
What it does
ExtracText for Confluence Cloud extracts all text from images and adds it to confluence search index making it available for searching. It also provides an advanced image editor which helps adding annotations to your images and all data is now contained within the image and available through search. A dedicated ExtracText Image search page allows to search and shows the images in thumbnails to identify them easily. There is also a macro that lets you search and add an existing image using text into any pages.
How I built it
We had the backend OCR technology built using Java for a while which we are using in "ExtracText for Confluence" server version. Now we are using the same technology in the backend and using Atlassian Connect framework for the front end. All apps are hosted in Google cloud Platform and highly available and scalable architecture with monitoring and production ready systems.
Challenges I ran into
Understanding how AtlassinConnectFramework works was the biggest challenge. There are very limited number of resources online and lot of trail and error before we get it right.
Accomplishments that I'm proud of
What I learned
Pay attention to little details. If not we will end up wasting lots of time on minor issues.
What's next for ExtracText for Confluence Cloud
Our mission is to solve unsolved useful problems and we are going to make this app the best image editing, creating tool available in the market with features no others have
Instructions for testing
Have provided our production link. The app is pending submission. Please use a trial version to test it out or let us know if you need a token generated.