def run(classifier, train_size, test_size, short_report=False):
print(classifier)
print("training set size: ", train_size)
print("test set size: ", test_size)
runner = cf.ClassifierRunner(classifier)
runner.train(mnist.training_images(train_size), mnist.training_labels(train_size))
runner.run(mnist.test_images(test_size), mnist.test_labels(test_size))
print(runner.get_report_str(short=short_report))
# Size that digit images are normalised to
NORMALISED_SIZE = (20, 20)
class KnnNormSizeDeslant(knn.KnnDigitClassifier):
def train(self, images, labels):
super(KnnNormSizeDeslant, self).train(
self.preprocess_all(images), labels)
def classify(self, image):
return super(KnnNormSizeDeslant, self).classify(self.preprocess(image))
def preprocess(self, image):
desl = imgproc.deslant(image)
return imgproc.crop_to_outer_contour(desl, NORMALISED_SIZE)
def preprocess_all(self, images):
for image in images:
yield self.preprocess(image)
if __name__ == '__main__':
NUM_TRAINS = 100
NUM_TESTS = 100
runner = cs.ClassifierRunner(KnnNormSizeDeslant())
runner.train(mnist.training_images(NUM_TRAINS), mnist.training_labels(NUM_TRAINS))
runner.run(mnist.test_images(NUM_TESTS), mnist.test_labels(NUM_TESTS))
print(runner.get_report_str())
def run(classifier, train_size, test_size):
print(classifier)
runner = cf.ClassifierRunner(classifier)
runner.train(mnist.training_images(train_size), mnist.training_labels(train_size))
runner.run(mnist.test_images(test_size), mnist.test_labels(test_size))