def create_model(scenario):
n_layer = scenario['layer']
n_column = get_total_columns(n_layer)
column_indexes = get_feature_columns(n_layer)
target_directory = get_feature_array_scenario_path(scenario['codename'])
target_files = get_files(target_directory)
target_train_files = target_files[train_start_index: train_end_index]
target_test_files = target_files[test_start_index: test_end_index]
X = np.zeros((0,n_column))
X_test = np.zeros((0, n_column))
for target_file in target_train_files:
# print target_directory+target_file
entry = np.load(target_directory+target_file)
X = np.concatenate((X,entry), axis = 0 )
for target_file in target_test_files:
# print target_directory+target_file
entry = np.load(target_directory+target_file)
X_test = np.concatenate((X_test,entry), axis = 0 )
Y = X[:, -1]
X = X[:,column_indexes]
Y_test = X_test[:, -1]
X_test = X_test[:,column_indexes]
strat_kfold = StratifiedKFold(Y, n_folds = 5)
# print "Train data: test data = %s : %s"%(len(X), len(X_test))
# clf1 = svm.SVC(class_weight='auto')
# clf2 = DecisionTreeClassifier()
clf3 = RandomForestClassifier()
def benchmark(classifier, X, Y, train_indices, test_indices):
classifier.fit(X[train_indices], Y[train_indices])
Y_predict = classifier.predict(X[test_indices])
return accuracy_score(Y[test_indices], Y_predict), confusion_matrix(Y[test_indices], Y_predict)
highest_accuracy = 0
highest_counter = 0
chosen_train_index = []
chosen_test_index = []
counter = 0
for train_index, test_index in strat_kfold:
acc, conf = benchmark(clf3, X, Y, train_index, test_index)
if acc > highest_accuracy:
highest_accuracy = acc
highest_counter = counter
chosen_train_index = train_index
chosen_test_index = test_index
counter += 1
# new_entry = {'acc':acc, 'fn':conf[0,1], 'fp':conf[1,0]}
# str_new_entry = str(new_entry) + " -- train-test:(%s,%s)"%(len(train_index), len(test_index))
# print str_new_entry
print "Chosen index-accuracy: %s -- %s"%(highest_counter, highest_accuracy)
clf3.fit(X[chosen_train_index], Y[chosen_train_index])
Y_predict = clf3.predict(X_test)
a_score, c_matrix = accuracy_score(Y_test, Y_predict), confusion_matrix(Y_test, Y_predict)
final_result = {'codename': scenario['codename'],'acc':a_score, 'fn':c_matrix[0,1], 'fp':c_matrix[1,0], 'tn':c_matrix[0,0], 'tp':c_matrix[1,1]}
str_final_result = "Test result --> " + str(final_result) + " test case:"+str(len(X_test))
print str_final_result
classifier_directory = classifier_path_px + scenario['codename'] + "/"
if not exists(classifier_directory):
makedirs(classifier_directory)
joblib.dump(clf3, classifier_directory+'RF'+"-"+category+"-"+scenario['codename']+ classifier_extension)
return final_result
def extract_feature(scenario):
n_layer = scenario['layer']
target_directory = get_feature_array_scenario_path(scenario['codename'])
create_directory(target_directory)
array_px_files = get_files(target_directory)
# Jangan lakukan ekstraksi fitur ulang
if len(array_px_files) >= 50:
print "feature "+scenario['codename']+" is already existed. Abort mission"
return
# Ambil semua file gambar
image_filenames = get_files(directory_path)
counter = 0
for image_filename in image_filenames:
# print "Extracting %s:%s"%(counter, position_file)
counter += 1
a = read_image(image_filename)
gt = read_groundtruth_image(image_filename)
# konversi menjadi binary image
gt = gt > 20
gt = gt.astype(int)
image_shape = a.shape
image_row = image_shape[0]
image_col = image_shape[1]
image_layer = image_shape[2]
im_slic = []
im_disp = []
im_bound = []
features = []
# Extract superpixel feature for each layer
for i in range(n_layer):
im_slic.append(slic(a, compactness=scenario['settings'][i]['compactness'],
n_segments=scenario['settings'][i]['segment'],
sigma=scenario['settings'][i]['sigma']))
im_slic[i] = label(im_slic[i], neighbors=8)
im_disp.append(np.copy(im_slic[i]))
im_bound.append(mark_boundaries(a, im_slic[i]))
temp_feature = regionprops(im_slic[i], intensity_image=rgb2gray(a))
features.append(list_to_dict(temp_feature))
X_indiv = []
for im_row in range(image_row):
for im_col in range(image_col):
# extract position and corresponding labels
posLabel = gt[im_row, im_col]
current_labels = []
# validate labels. 0 label is not allowed. causing not exists error
valid_position = True
for i in range(n_layer):
current_level_labels = im_slic[i][im_row, im_col]
current_labels.append(current_level_labels)
if current_level_labels == 0:
valid_position = False
break
if not valid_position:
continue
# concat all layer properties
x_entry = []
for i in range(n_layer):
feat = features[i][current_labels[i]]
for att in attributes:
if att == 'bbox':
(min_row, min_col, max_row, max_col) = feat['bbox']
x_entry.append(min_row)
x_entry.append(min_col)
x_entry.append(max_row)
x_entry.append(max_col)
else:
x_entry.append(feat[att])
if posLabel == 1:
mark(current_labels[i], 1, im_slic[i], im_disp[i])
x_entry.append(posLabel)
X_indiv.append(x_entry)
f = get_feature_array_file(scenario['codename'], image_filename, mode='w')
X_indiv = np.array(X_indiv)
X_indiv_u = unique_rows(X_indiv)
np.save(f, X_indiv_u)
f.close()
@author: fruity
'''
import numpy as np
from sklearn.metrics import confusion_matrix, accuracy_score
from constant import array_px_path, train_start_index, train_end_index, test_start_index, test_end_index, classifier_path_px, classifier_extension, category
from os import listdir, makedirs
from os.path import isfile, join, splitext, exists
from sklearn.ensemble import RandomForestClassifier
from sklearn.externals import joblib
from sklearn.cross_validation import StratifiedKFold, cross_val_score
from util import get_feature_columns, get_total_columns, get_feature_array_scenario_path
from util.file import get_files
target_path = get_feature_array_scenario_path("base3")
print target_path
def create_model(scenario):
n_layer = scenario['layer']
n_column = get_total_columns(n_layer)
column_indexes = get_feature_columns(n_layer)
target_directory = get_feature_array_scenario_path(scenario['codename'])
target_files = get_files(target_directory)
target_train_files = target_files[train_start_index: train_end_index]
target_test_files = target_files[test_start_index: test_end_index]
X = np.zeros((0,n_column))
X_test = np.zeros((0, n_column))
