Currently our feature vector uses counts of the number of each distinct color in each image. This works reasonably well on our small data set, but it's not totally accurate. See https://github.com/zeke/github-avatars/pull/2
For our next session let's create a new feature vector using:
Currently our feature vector uses counts of the number of each distinct color in each image. This works reasonably well on our small data set, but it's not totally accurate. See https://github.com/zeke/github-avatars/pull/2
For our next session let's create a new feature vector using:
cc @andreasjansson