All three of us working on the project love traveling and want to fly to new places in the world on a budget. By combining Google computer vision to recognize interesting places to go as well as JetBlue's flight deals to find the best flights- we hope we've created a product that college students can use to explore the world!
What it does
Envision parses through each picture on websites the user is on and predicts a destination airport from the image given entities tracked through computer vision. It finds the best JetBlue flight deal based on current location, price, and similarity in destination and returns a hover over chrome extension best deal recommendation. It links to a website that shows more information about the flight, including flight type, fare type, etc, as well as pictures of the destination found through Google Places API. Envision travel effortlessly today! :))
How we built it
We build Envision using JetBlue's Deals data and the Google Cloud Platform- Google Vision API, Google Places API. First, we scraped images from Google Image Search of every JetBlue airport location. Running every image through Google Vision API, we received a list of entities that are found in the most common images. By creating a chrome extension to track images in every webpage, each picture is then translate through computer vision into an entity list, and that list is used to find the best location to recommend / most similar destination. Using JetBlue's Deals data, we found the best deal flight from the closest airport based on current location, and routed to our target airport destination, using Google Places API Place Nearby Search, Text Search, Place Details combined to find the most suitable flight to take. Using Heroku and Flask to host our web services, we created a workflow that leads to a React website with more information about the recommended flights and photos of the destination similar to that of the original images on the browsed website.
Challenges we ran into
There are many steps in our processing pipeline- which includes receiving images from the chrome extension, parsing the image to an entity list on Google Cloud Vision, finding a recommended location and the best JetBlue flight that leads to a region, as well as similar images from the area that links to the original image shown in a separate website. To connect every part together through endpoints took a lot of figuring out near the end!
Accomplishments that we're proud of
Creating a working product!
What we learned
Lots of web & API calling.
What's next for Envision
Creating a more user intuitive interface for Envision on the chrome extension as well as the website.