def test_direction_cone_of_circle_west(debug=False):
circle = Point(0, 0).buffer(1)
cone = ops.direction_cone(circle, (0, -1), pi / 2, 5)
assert Point(0.5, -1).within(cone)
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, circle, color='lightcoral')
plot_utils.plot(axes, cone, color='lightblue', marker='x')
plt.show()
def test_direction_cone_of_triangle_south(debug=False):
triangle = Polygon([(0, 0), (0.5, 1), (1, 0)])
cone = ops.direction_cone(triangle, (0, -1), pi / 2, 2)
assert Point(0.5, -1).within(cone)
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, triangle, color='lightcoral')
plot_utils.plot(axes, cone, color='lightblue', marker='x')
plt.show()
def test_outer_inner_buffer_square(debug=False):
square = Polygon([(-1, -1), (-1, 1), (1, 1), (1, -1)])
oi_buffer = ops.outer_inner_buffer(square, 0.2)
assert Point(0.9, 0.9).within(oi_buffer)
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, square, color='lightgrey', marker='x')
plot_utils.plot(axes, oi_buffer, color='lightcoral', marker='+')
plt.show()
def test_outer_inner_buffer_circle(debug=False):
circle = Point(0, 0).buffer(1)
oi_buffer = ops.outer_inner_buffer(circle, 0.2)
assert Point(0.49 * 2**0.5, 0.49 * 2**0.5).within(oi_buffer)
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, circle, color='lightgrey', marker='x')
plot_utils.plot(axes, oi_buffer, color='lightcoral', marker='+')
plt.show()
def test_medial_line_of_polygon_3(debug=False):
polygon = Polygon([(1, 1), (2, 1.5), (0, 3), (-1, 4), (-2, 3), (-0.5, 2)])
medial_line = ops.medial_line(polygon)
assert medial_line.within(polygon)
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, polygon, color='lightgrey', marker='x')
plot_utils.plot(axes, medial_line, color='lightcoral')
plt.show()
def test_medial_line_of_polygon_2(debug=False):
polygon = Point(1, 1).buffer(2).union(Point(2, 2).buffer(2))
medial_line = ops.medial_line(polygon)
assert medial_line.within(polygon)
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, polygon, color='lightgrey', marker='x')
plot_utils.plot(axes, medial_line, color='lightcoral')
plt.show()
def test_medial_line_of_polygon_1(debug=False):
polygon = Polygon([(1, 1), (1, 1.5), (1, 2), (1, 3), (2, 3), (2, 2), (2, 1.5), (2, 1), (1.5, 1.5)])
medial_line = ops.medial_line(polygon)
assert medial_line.touches(Point(1.5, 2))
assert medial_line.touches(Point(1.25, 1.75))
assert medial_line.touches(Point(1.5, 2.5))
assert medial_line.touches(Point(1.5, 2))
assert medial_line.touches(Point(1.75, 1.75))
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, polygon, color='lightgrey', marker='x')
plot_utils.plot(axes, medial_line, color='lightcoral')
plt.show()
def test_between_triangles(debug=False):
left = Polygon([(1, 1), (2, 2), (3, 1)])
right = Polygon([(4, 2), (5, 1), (6, 2)])
between_result = between(left, right)
assert between_result.equals(Polygon([(2, 2), (3, 1), (5, 1), (4, 2)]))
if debug:
plt.figure()
axes = plt.subplot(111)
plot_utils.plot(axes, left, color='lightcoral')
plot_utils.plot(axes, right, color='lightblue')
plot_utils.plot(axes, between_result, color='lightgrey')
plt.show()
# cols_to_merge_with_name = ["threads_per_datapoint", "datapoints_per_block"]
cols_to_merge_with_name = ["datapoints_per_block"]
col_x_axis = "threads_per_datapoint"
cols_y_axis = ["train time"]
for col_y_axis in cols_y_axis:
lines = {}
# print(col_y_axis)
for idx, _ in enumerate(csv["name"]):
line_name = plot_utils.get_line_name(idx, csv, cols_to_merge_with_name)
if line_name not in lines:
lines[line_name] = {'x': [], 'y': []}
lines[line_name]['x'].append(csv[col_x_axis][idx])
lines[line_name]['y'].append(csv[col_y_axis][idx])
plot_utils.adapt_baselines(lines)
ax = plt.subplot(1, 1, 1)
ax.set_xscale('log', basex=2)
plot_utils.plot(lines, col_x_axis, col_y_axis, ax)
plt.show(block=False)
raw_input("press enter to close")
plt.close("all")
y=labels,
batch_size=batch_size,
num_epochs=1,
shuffle=False)
eval_results = mnist_classifier.evaluate(input_fn=eval_input_fn)
accuracy = eval_results["accuracy"]
if eval_results["accuracy"] < 0.95:
raise ValueError("Accuracy is {:5}% which is low".format(accuracy*100))
# Generate adversarial images
images, adv_images = generate_adversarial_images(
mnist_classifier,
images, labels,
old_label, new_label,
n_to_modify, reg_lambda,
one_pixel)
# Infer the classes
classes = predict(mnist_classifier, images)
adv_classes = predict(mnist_classifier, adv_images)
# Compute the deltas
deltas = adv_images - images
# Plot and save the figure
fig = plot(images, deltas, adv_images, classes, adv_classes)
if not os.path.exists('images/'):
os.mkdir('images/')
fig.savefig('images/example.png' if not one_pixel else 'images/one_pixel_example.png')
out_accuracy_model_file = os.path.abspath(os.path.join(out_dir, 'model_accuracy.png')) out_recall_file = os.path.abspath(os.path.join(out_dir, 'positive_recall.png')) neg_recall_file = os.path.abspath(os.path.join(out_dir, 'negative_recall.png')) out_fscore_file = os.path.abspath(os.path.join(out_dir, 'positive_fscore.png')) neg_fscore_file = os.path.abspath(os.path.join(out_dir, 'negative_fscore.png')) out_acc_file = os.path.abspath(os.path.join(out_dir, 'accuracy.png')) pos_precision_acc = numpy.mean(metrics_pos_precision_scores, axis=0) # prec_y_data_list.append(pos_precision_acc) # prec_y_data_list plot(epochs_arr, metrics_pos_precision_scores, 'Positive Precision', 'Epoch', 'Precision', epochs_num, out_prec_file, lw_lst, alpha_lst, fold_pos_prec_mean) plot(epochs_arr, metrics_neg_precision_scores, 'Negative Precision', 'Epoch', 'Precision', epochs_num, neg_prec_file, lw_lst, alpha_lst, fold_neg_prec_mean) plot(epochs_arr, metrics_acc_scores, 'Accuracy', 'Epoch', 'Accuracy', epochs_num, out_acc_file, lw_lst, alpha_lst, fold_accuracy_mean, [70, 101]) plot(epochs_arr, metrics_pos_recall_scores, 'Positive Recall', 'Epoch', 'Recall', epochs_num, out_recall_file, lw_lst, alpha_lst, fold_pos_recall_mean) plot(epochs_arr, metrics_neg_recall_scores, 'Negative Recall', 'Epoch', 'Recall', epochs_num, neg_recall_file, lw_lst, alpha_lst, fold_neg_recall_mean) plot(epochs_arr, metrics_pos_fscore_scores, 'Positive F-score', 'Epoch', 'F-score', epochs_num, out_fscore_file, lw_lst, alpha_lst, fold_pos_fscore_mean) plot(epochs_arr, metrics_neg_fscore_scores, 'Negative F-score', 'Epoch', 'F-score', epochs_num, neg_fscore_file, lw_lst, alpha_lst, fold_neg_fscore_mean)
cols_to_merge_with_name = ["datapoints_per_block"]
col_x_axis = "threads_per_datapoint"
cols_y_axis = ["train time"]
for col_y_axis in cols_y_axis:
lines = {}
# print(col_y_axis)
for idx, _ in enumerate(csv["name"]):
line_name = plot_utils.get_line_name(idx, csv, cols_to_merge_with_name)
if line_name not in lines:
lines[line_name] = {'x':[], 'y':[]}
lines[line_name]['x'].append(csv[col_x_axis][idx])
lines[line_name]['y'].append(csv[col_y_axis][idx])
plot_utils.adapt_baselines(lines)
ax = plt.subplot(1, 1, 1)
ax.set_xscale('log', basex=2)
plot_utils.plot(lines, col_x_axis, col_y_axis, ax)
plt.show(block=False)
raw_input("press enter to close")
plt.close("all")
def plot(self, start=0, time=None, **plot_kwargs):
if time is None:
fin = len(self.array)
else:
fin = start + time * self.rate
plot(self.array[start:fin], **plot_kwargs)