Motivated to use OpenCV, our team set out to create an application that would generate restaurant reviews based on a picture of its logo, while providing insight about the nutritional value about its food. Ultimately, our objective was to help create a healthy and food-conscious society.
What it does
The user snaps a photo of a restaurant's logo, and sends it to our server using Twilio's API. Then, our server uses Google's Cloud Vision API to determine what restaurant the logo belongs to. Our server proceeds by querying the Google Places API to obtain reviews about the restaurant. These reviews are then displayed on a corresponding webpage.
How we built it
We used Twilio's API to interface between the user's phone and our server. We also used Google's Cloud Vision and Places API's to get reviews about the restaurant. We incorporated a MongoDB database to store backlogs of reviews to reduce the number of external queries.
Challenges we ran into
We had originally planned to use OpenCV to do our own logo and text-recognition. However, our attempts proved unfruitful. This ultimately led us to use Google's Cloud Vision API. In general, the lack of time along with unfamiliarity with the API's prevented us from implementing all of the features we had originally planned to incorporate.
Accomplishments that we're proud of
Though they did not appear in our final product, we implemented various text-recognition algorithms (such as feature-matching and k-Nearest Neighbors).
What we learned
Working on this project gave us a reason to learn the basics of machine learning (particularly its application towards computer vision). Moreover, we were able to gain a deeper understanding of how OpenCV works under-the-hood. We were also able to learn how to use Twilio and Google's API's.
What's next for AskHer
We plan to integrate our own text-recognition algorithms in the future to become less dependent on Google's API. Another goal that we have for AskHer is to create a Profile system, which would allow users to customize their searches based on their dietary restrictions. This would also allow users with certain diseases to see if a restaurant suits their condition. As it is to be expected, we would like to incorporate the rest of the features that were not implemented during HackRU.