algolia / color-extractor

Extract the dominant color(s) of your fashion articles!
https://algolia.com
Other
277 stars 69 forks source link

No 'white' color in the dataset #12

Open kHarshit opened 4 years ago

kHarshit commented 4 years ago

The xkcd dataset (color_names.npz) used for KNN to predict color name from RGB values doesn't has white color.

Counter({'black': 1782,
         'green': 53000,
         'blue': 54785,
         'brown': 10529,
         'purple': 26416,
         'maroon': 3283,
         'red': 15476,
         'pink': 13617,
         'orange': 9152,
         'yellow': 7857,
         'mustard': 711})

In fact, it has only a few colors, not all 27 as mentioned on the blog post.

We turned to a K-Nearest-Neighbors algorithm to give color names to RGB values, thanks to the XKCD Color Survey. The XKCD survey consists of 200,000 RGB values labeled with 27 different color names (e.g. black, green, teal, etc.) that we use to train a scikit-learn KNeighborsClassifier.

How do I add white to dataset as I need it.

drewburnett commented 4 years ago

Any progress?

Haroenv commented 4 years ago

This project is released as-is, and no longer maintained by us, sorry! I'll add a note to the readme to make that clearer.