The constructor

trainer = cv.BOWKMeansTrainer(dictionarySize)
[...] = cv.BOWKMeansTrainer(...,'OptionName', optionValue, ...)

Input

• dictionarySize Number of clusters.

Options

• Criteria The algorithm termination criteria, that is, the maximum number of iterations and/or the desired accuracy. The accuracy is specified as Criteria.epsilon. As soon as each of the cluster centers moves by less than Criteria.epsilon on some iteration, the algorithm stops. default struct('type','Count+EPS', 'maxCount',100, 'epsilon',eps('float'))
• Attempts The number of times the algorithm is executed using different initial labelings. The algorithm returns the labels that yield the best compactness. default 3.
• Initialization Method to initialize seeds, default 'PP'. One of the followings:
  • 'Random' Select random initial centers in each attempt.
  • 'PP' Use kmeans++ center initialization by Arthur and Vassilvitskii [Arthur2007].