Travel Match

webpage showing vacation suggestion listings

Inspiration

Often times, we travel to destination based on its popularity like New York or Hawaii, rather than looking for places that we would really enjoy the most. So we decided to create a web application that would recommend travel destinations to users based on images they upload based on their own interests. Each suggestion is tailored to every query.

What it does

This website will generate a list of suggestions based off a user inputted image via url or upload. Using Microsoft Azure Computer Vision, the website will generate a list of suggested travel destinations, including lowest roundtrip airfare and predicted non-flight related vacation costs, by analyzing the tags from Computer Vision.

How we built it

We have three main documents for code- the database of locations, the html file for the website, and the javascript file that defines the methods and implements the API. The html website receives user input and returns the suggestions. The database contains formatted data about various possible travel destinations. The javascript file gathers information from each inputted photo, and assigns the photo scores in various categories. Those scores are then compared to those in the database, and using factors such as cost and location type, we generate a list of 7 travel recommendations.

Challenges we ran into

While trying to acquire a suitable database, we attempted to scrape Wikipedia articles using node.js. After using cheerio and building the basic structure, we found that the necessary information was located in different parts of each article. We were ultimately unable to write code that could filter through the article and extract information, and had to hand generate a database after 5 or more hours of work. Another challenge was finding efficient comparison method so our website did not take too long to process photos.

Accomplishments that we're proud of

Our website accomplished our main goal. It is able to take in an image, process the tags, and return a list of possible destinations. We also have a basic algorithm that maximizes the accuracy of our website given the current database format.

What we learned

Since our team had first-time hackers, who were also first time javascript coders, we learned a great deal about how the language functioned, and standards for coding website. We also learned how to share and update code through gitHub. More specifically to our project, this was the first time we used Microsoft Azure and Amadeus APIs, and learned how to implement them into our code.

What's next for Travel Match

We want to build a more comprehensive database, that includes more locations with more specific data. We also want to come up with a more efficient and accurate comparison algorithm than we currently have.