We wanted to do cool data visualizations and decided that looking at color would lend itself to beautiful data viz.

What it does

We pulled in 70,000+ records of artwork from the Tate Collection. We then sent the images to Clarifai to get the colors of the images. Then we used D3.js to visualize trends in the data.

How we built it

Imported 70k records into a MongoDB instance. Filtered by data with non-empty thumbnailUrl fields. Processed records by appending the colors returned from the Clarifai API. Queried results and processed data with MongoDB and Python. Exported to JSON. Rendered JSON with D3.js. Developed and served on a Linode server.

Challenges we ran into

Running all records at once broke every now and then. D3 is hard. A lot of art pieces were in greyscale. We hadn't used any of these technologies before (besides Python).

Accomplishments that we're proud of

We hadn't used any of these technologies before. The machine learning from Clarifai worked great. We actually got results in a short time.

What we learned

We learned that doing data analysis on something you're familiar with is probably easier than something you have no idea about (art history). Machine learning is cool.

What's next for ColorTrends

More data analysis!

Share this project: