all_accuracies_list = []
all_labels = []
for current_classifier in selected_classifiers:
print(
"~ Running program with {} classifier.".format(current_classifier))
classifier = Classifier(current_classifier)
accuracy_list = []
start_time = time.time()
for i, (train, test) in enumerate(
kfold_cross_validation(data, k=int(args["kfold"]))):
samples, labels = parse_samples_labels(train)
classifier.fit(samples, labels)
samples, labels = parse_samples_labels(test)
accuracy = classifier.accuracy(samples, labels)
if args["verbose"]:
print("+ [{:02d}/10] accuracy: {:3d}%".format(
i + 1, int(accuracy * 100)))
if draw_confusion_matrix:
confusion_matrix = classifier.confusion_matrix(samples, labels)
accuracy_list.append(accuracy)
all_accuracies_list.append(accuracy_list)
all_labels.append(current_classifier)
if args["verbose"]:
print("- Took {:.2f} seconds.".format(time.time() - start_time))
data = DataProvider(cars_paths,
non_cars_paths,
car_id=car_id,
non_car_id=non_car_id,
test_size=0.2)
data.save_labels_info_graph()
data.save_random_images(car_id, "../training_results/random_car_images.png")
data.save_random_images(non_car_id,
"../training_results/random_non_car_images.png")
print("Training is in progress...")
clf.train(data.train.features, data.train.labels)
print("Training time: {:.2f} min.".format(clf.training_time / 60))
accuracy = clf.accuracy(data.test.features, data.test.labels)
print("Test accuracy: {:.3f}%.".format(accuracy * 100))
space = "-" * 50
info = [
"Accuracy: {:.4f}%".format(accuracy * 100), space,
"\t- Features vector compound parts -", "features_providers: [{}]".format(
", ".join([type(i).__name__ for i in features_providers
])), "color_space: {}".format(color_space),
"classifier_image_size: {}".format(classifier_image_size), space,
"\t- Spatial binning of color parameters -",
"spatial_binning_size: {}".format(spatial_binning_size), space,
"\t- Color histogram parameters -",
"color_histogram_n_bins: {}".format(color_histogram_n_bins),
"bins_range: {}".format(bins_range), space,
"\t- Histogram of oriented gradients parameters -",
