mlp_training.py

__author__ = 'zhengwang'

import cv2
import numpy as np
import glob
import sys
from sklearn.model_selection import train_test_split

print 'Loading training data...'
e0 = cv2.getTickCount()

# load training data
image_array = np.zeros((1, 38400))
label_array = np.zeros((1, 4), 'float')
training_data = glob.glob('training_data/*.npz')

# if no data, exit
if not training_data:
    print "No training data in directory, exit"
    sys.exit()

for single_npz in training_data:
    with np.load(single_npz) as data:
        train_temp = data['train']
        train_labels_temp = data['train_labels']
    image_array = np.vstack((image_array, train_temp))
    label_array = np.vstack((label_array, train_labels_temp))

X = image_array[1:, :]
y = label_array[1:, :]
print 'Image array shape: ', X.shape
print 'Label array shape: ', y.shape

e00 = cv2.getTickCount()
time0 = (e00 - e0)/ cv2.getTickFrequency()
print 'Loading image duration:', time0

# train test split, 7:3
train, test, train_labels, test_labels = train_test_split(X, y, test_size=0.3)

# set start time
e1 = cv2.getTickCount()

# create MLP
layer_sizes = np.int32([38400, 32, 4])
model = cv2.ANN_MLP()
model.create(layer_sizes)
criteria = (cv2.TERM_CRITERIA_COUNT | cv2.TERM_CRITERIA_EPS, 500, 0.0001)
#criteria2 = (cv2.TERM_CRITERIA_COUNT, 100, 0.001)
params = dict(term_crit = criteria,
               train_method = cv2.ANN_MLP_TRAIN_PARAMS_BACKPROP,
               bp_dw_scale = 0.001,
               bp_moment_scale = 0.0 )

print 'Training MLP ...'
num_iter = model.train(train, train_labels, None, params = params)

# set end time
e2 = cv2.getTickCount()
time = (e2 - e1)/cv2.getTickFrequency()
print 'Training duration:', time
#print 'Ran for %d iterations' % num_iter

# train data
ret_0, resp_0 = model.predict(train)
prediction_0 = resp_0.argmax(-1)
true_labels_0 = train_labels.argmax(-1)

train_rate = np.mean(prediction_0 == true_labels_0)
print 'Train accuracy: ', "{0:.2f}%".format(train_rate * 100)

# test data
ret_1, resp_1 = model.predict(test)
prediction_1 = resp_1.argmax(-1)
true_labels_1 = test_labels.argmax(-1)

test_rate = np.mean(prediction_1 == true_labels_1)
print 'Test accuracy: ', "{0:.2f}%".format(test_rate * 100)

# save model
model.save('mlp_xml/mlp.xml')