Created: 2014-05-18 19:48
Updated: 2018-03-12 04:47
License: mit


Experimental word classifier based on dynamic time warping

Pleiad helps you identify words in an image. Use it when you dont need extensive character recognition and/or dont have much data, and need to identify few pre planned words only.

Pleiad doesn't need extensive training data, just a single image for each class (few more for better result) is all you need.


  • Install R (and python)
  • install.packages('dtw') in R shell
  • pip install -r requirements


  • Crop words from image and stretch to a fixed size (one size per classifier)


  • Create Word objects from word images

from pleiad import pleaid
word = pleiad.Word(image, "climb")
  • Train classifier from a list of Words
classifier = pleiad.PleiadClassifier(image.shape)
  • Predict
  • Save for future use'classifierOne')


Pleiad works by treating the outer outline of each image of word as a time series and predicting by using the dynamic time warping distance between the series.



Copyright (c) 2014 Abhinav Tushar

Cookies help us deliver our services. By using our services, you agree to our use of cookies Learn more