Planėt

Home page
Login page
User picks which path they would like to take
Select image
Location Results
WRLD 3D Map of location

Inspiration

Humans usually only know what they like once they see it. The same holds true for travel destination. Planėt helps users plan their trips around the world by learning their aesthetic preferences through a quick, image based survey. Using Computer Vision to classify images and Machine Learning to learn user preferences, Planėt is the new go to solution for quick travel inspiration.

What it does

Planėt is an interactive Web App where users are repeatedly asked to pick the travel picture that appeals most to them out of two images. Using Google's Cloud Machine Learning, the platform analyzes images from the Bing Image API and categorizes them using advanced computer vision. With the resulting data, Planėt is able to match the user with preferences it learns through a quick, dual paned survey. Planėt analyzes this information using a decision tree and narrows the user's travel options to five recommendations.

How we built it

Planėt was built with Flask, a Python Web Framework, an SQL Database, and a front end with Bootstrap, CSS, HTML, and a hint of Javascript. In regards to the Machine Learning, the images were collected through the Bing API, and analyzed with Google's Cloud Machine Learning Platform. To learn the user's destination results, we used the Google Computer Vision API to analyze images related to a variety of destinations. After multiple rounds of classification, we came up with a precise set of attributes for each destination, that can be matched to the preferences learned during the image processing survey.

Challenges we ran into

We developed the app on a variety of testing platforms, but struggled with putting the final product together. The website itself had some issues with the integration to the machine learning platform but we are currently working through it in the final stretch.

Accomplishments that we're proud of

For all of our team, this was our first web development hack, as well as our first ML hack. We decided to try something that we hadn't done before and although it may not have worked out perfectly, it definitely was beyond what we expected when we came into Cal Hacks this weekend.

What's next for Planėt

An extension to the application is to use the same algorithms to make decisions on food preferences and suggest restaurants from a visual analysis, and we plan on integrating this to our main application to fulfill the vision of truly Planning It.