Semantically-resonant color assignments (such as mapping "ocean" to "blue") are useful in categorical visualizations, and can facilitate value lookup and comparisons.

This repository contains C# code that assigns a set of categorical concept names to semantically-resonant colors.
The algorithm makes use of images to determine frequent and discirminative colors for each concept name. It uses Google Image Search to query for relevant images. Project page: link

Main file: SemanticColors.sln

• [c3_data.json] (https://github.com/StanfordHCI/c3/tree/master/data/xkcd) - Needed for computing color name distances. Download to some place on your computer
• [Google CustomSearch C# Library v1.3.0] (https://code.google.com/p/google-api-dotnet-client/wiki/APIs#CustomSearch_API) - Add/Update the reference to Google.Apis.dll and Google.Apis.Customsearch.v1.dll to Engine.csproj
• You'll also need to sign up for an API key and set up a CustomSearch engine. There's a daily query limit for the free option.
• Create a new search engine with an initial site url.
• Then edit the engine, and remove the site from Sites to search. Change the selected search option to Search the entire web but emphasize included sites.
• Enable Image Search.

The code has been tested on Windows 7.

Usage: SemanticColors [inFile] [outFile] [paramFile] [renderDir]

• inFile - file with list of categories to assign colors. See sampleCategories.csv for an example
• outFile - output file where color assignments are saved
• paramFile - config file. See sampleParams.txt for an example where
• imageDir - is where images of concept names will be saved
• cacheDir - is where computed histograms will be saved and re-used
• jsonFile - is the location of the c3_data.json file
• apiKey - is your Google CustomSearch API key
• cx - is your Google CustomSearch engine id
• renderDir (optional)- If included, images of the color assignments and affinity scores for each category are saved in this directory.