I created an approach and code project that could read colors in text and standardize them into primary colors, color families, and hexadecimal color codes to enable ecommerce product listings to be categorized, listed, filtered and searched by colors without access to product images from manufacturers. This was comprised of
- converting 30K+ colors into their values in RGB/HSL spaces
- measuring adjectives affected their color values (e.g. “light”, “deep”, “bright”, “soft”, etc)
- grouping them into color families RGB/HSL spaces with density based clustering
- creating color group names using a combination of corpus-frequencies and RGB centroids
