To explore the concept of ML. To become familiar with the concept of image recognition, and OCR.
What it does
DeCom is a mobile cross platform app that enables users to incorporate image recognition capabilities. To be exact, that is OCR (Optical Character Recognition) and object detection. For both outputs the user is able to either open a link(if applicable) or run a valid search query.
How we built it
DeCom is written using XCode and Android Studio, using java, MongoDB, Cloud and Native apis, objective-c and Swift.
Challenges we ran into
The challenge was to improve OCR accuracy, Object detection accuracy and implement features for searching and opening links. For the iOS platform the challenge was configuring the Firebase cloud platform, as we were unfamiliar with the task.
Accomplishments that we're proud of
We managed to obtain a trained model for object detection for iOS and also link the Firebase google cloud Tensor flow api to highly increase our OCR accuracy. Furthermore, for Android we were able to implement text translation using Google translate api.
What we learned
We never used Xcode! So it was a blast for us to make an iOS app. Our app intended to have a server side, incorporating mongoDB, so it was interesting to work with the database using python to store and query information. Finally, Google's Firebase and Tensor-flow apis, and CoreML were other useful libraries that we enjoyed using for the first time.
What's next for DeCom
Further improve the OCR and Object detection. Further incorporate cloud platforms, and link the server side (MongoDB) to improve the versatility of our platform.