They say photos are worth a thousand words. We wanted to put those words to use. These words, or tags, have become an important part of the digital world, as we rely on photo or video tags for image discovery, categories, inspiration, and search engine optimization. In many cases, marketing to target markets with unique demographics in certain regions of the world has been difficult. With GeoTrends, tags can be developed automatically and accurately.

What it does

GeoTrends take recent uploads from Flickr, as popular photography website, and uses machine learning software from Clarifai to develop relevant image and subject identification. Geographical data places these points on a map with ESRI's ArcGIS platform.

How we built it

Our application was built as a website. The majority of the logic is written in javascript. In the future, a more robust web application architecture will make more comprehensive use of the data.

Challenges we ran into

The Asynchronous nature of JavaScript made it difficult to ensure that the keywords associated with a particular picture and location actually populated the map correctly (having the correct keywords refer to the correct image)

Accomplishments that we are proud of

Making the keywords properly refer to the photos they were associated with was a milestone considering JavaScript's asynchronous nature. In addition, calling information from three different API's was an incredible accomplishment.

What we learned

We learned that it was very difficult to learn an API in less than 24 hours. We have barely scratched the surface with the ESRI API, and I know that there are some great tools that can enrich this application.

What's next for GeoTrends

For every photo uploaded, tags relevant to the photos are automatically generated alongside it - providing ease of use and simplicity to experienced or inexperienced users. Clarifai's "always learning" algorithms will ensure GeoTrends will keep a competitive advantage in image subject identification while ESRI will give us an edge in displaying trend data analytics. The effective combination of the two technologies will be the role model for the tools in trend research, trend analytics, and trend prediction for target markets of all characteristics based on past and real-time data.

Brought to you by: Phillip Sloan, Eric Hanamoto, Andrew Larke, and Emil Angelo Rodriguez

Built With

Share this project: