ISI - Incremental Splitting of Intervals

Hirata, Nina ST, Roberto Hirata Jr, and Junior Barrera. "Basis computation algorithms." Mathematical Morphhology and its Applications to Signal and Image Processing (Proceedings of the 8th International Symposium on Mathematical Morphology). 2007.

ISI (short for Incremental Splitting of Intervals) is a classifier specifically designed for image transform learning. It is based on mathematical morphology operators and can be decomposed as a combination of simple image transformations.

This implementation has several limitations: (i) only works with binary inputs; (ii) has high memory and CPU consumption and (iii) uses a specific encoding for features, so it can not be coupled with other methods in TRIOS.

According to some simple tests, ISI does achieve performance slightly better than using Decision Trees from scikit-learn. However, ISI is much more expensive in terms of CPU and memory when larger (more than 40) sets of features are used.

The code below illustrates how to use ISI to process the jung dataset. ISI only works with RAWBitFeatureExtractor and has no parameters. The code below takes about a minute to run. The same approach using a Decision Tree takes less than 10 seconds and results in very similar performance.

Only use ISI with small numbers of features. Windows larger than 40 points take forever to run.

# file docs/examples/methods/isi.py
from trios.classifiers.isi import ISI
from trios.feature_extractors import RAWBitFeatureExtractor
import trios

import trios.shortcuts.window as trios_win

if __name__ == '__main__':
    training_set = trios.Imageset.read('../jung-images/training.set')
    test = trios.Imageset.read('../jung-images/test.set')        
    win = trios_win.rectangle(7, 7)

    # ISI requires features coded in the bits of an int32
    op = trios.WOperator(win, ISI, RAWBitFeatureExtractor)
    op.train(training_set)
    print('Error: ', op.eval(test))

Error:  0.013615260395639888