def mlp():
from sklearn.neural_network import MLPClassifier
# Prepare X_train, X_test
X_train_names, y_train, X_test_names, y_test = prepare_dataset()
X_train = []
X_test = []
signature_length = 200
for file_path in X_train_names:
im = cv2.imread(file_path, cv2.IMREAD_GRAYSCALE).astype("uint8")
signature, _ = signature_descriptor(im, signature_length)
X_train.append(signature)
for file_path in X_test_names:
im = cv2.imread(file_path, cv2.IMREAD_GRAYSCALE).astype("uint8")
signature, _ = signature_descriptor(im, signature_length)
X_test.append(signature)
# Create MLP
clf = MLPClassifier(random_state=1,
max_iter=300,
hidden_layer_sizes=(100, 50)).fit(X_train, y_train)
print("MLP score", clf.score(X_test, y_test))
# Confusion matrix
from sklearn.metrics import plot_confusion_matrix
import matplotlib.pyplot as plt
plot_confusion_matrix(clf, X_test, y_test)
plt.xticks(rotation=90)
plt.show()
def main():
X_train, y_train, X_test, y_test = prepare_dataset()
name = 3
file = f"./dataset/tests/{name}.png"
file = X_train[8]
img = cv2.imread(file, cv2.IMREAD_GRAYSCALE).astype("uint8")
unl_fourier(img, 4)
cv2.imshow("Original image", img)
cv2.imshow("UNL", unl(img))
cv2.waitKey(0)
def test():
X_train, y_train, X_test, y_test = prepare_dataset()
im = cv2.imread(X_train[8], cv2.IMREAD_GRAYSCALE).astype("uint8")
contour = object_contour(im)
# plt.imshow(im)
# plt.show()
# Center of contour
x, y = center_of_contour(im, contour)
# Plot image with center
cv2.circle(im, (x, y), 7, (255, 255, 255), -1)
cv2.imshow("Image", im)
cv2.waitKey(0)
# Contours, what are those?
# Contour points start with first upper left point
# for idx, point in enumerate(contour):
# print(point)
# cv2.circle(im, (point[0, 0], point[0, 1]), 7, (25 * idx, 25 * idx, 25 * idx), -1)
# if idx > 10:
# break
# cv2.imshow("Image", im)
# cv2.waitKey(0)
# Calculate distance to every point
dists = calculate_dists(contour, [x, y])
# Plot signature
# print(dists)
# plt.plot(dists)
# plt.show()
# Interpolate dists vector to fixed length e.g. 200 elements
downsampled_y, downsampled_x = scale_dists_length(dists, 200)
# Plot points
plt.plot(dists)
plt.scatter(downsampled_x, downsampled_y)
plt.show()
# Downsampled points length
print("Downsampled distances lentgh:", downsampled_y.shape)
print("Distances lentgh:", dists.shape)
plt.plot(downsampled_y)
plt.show()
def main():
from simple_shape_descriptors import prepare_dataset
X_train, y_train, X_test, y_test = prepare_dataset()
sizes = [5, 20, 30, 35, 40]
for size in sizes:
clf = FourierClassifier(size=size)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
from sklearn.metrics import accuracy_score
acc = accuracy_score(y_test, y_pred)
print(f"Size: {size} Acc: {acc}")
def main():
X_train, y_train, X_test, y_test = prepare_dataset()
clf = SignatureClassifier(signature_length=200)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
from sklearn.metrics import accuracy_score
acc = accuracy_score(y_test, y_pred)
print(f"Acc: {acc}")
from sklearn.metrics import plot_confusion_matrix
import matplotlib.pyplot as plt
plot_confusion_matrix(clf, X_test, y_test)
plt.xticks(rotation=90)
plt.show()
def experiment_30():
from simple_shape_descriptors import prepare_dataset
X_train, y_train, X_test, y_test = prepare_dataset()
size = 30
clf = FourierClassifier(size=size)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
from sklearn.metrics import accuracy_score
acc = accuracy_score(y_test, y_pred)
print(f"Size: {size} Acc: {acc}")
from sklearn.metrics import plot_confusion_matrix
import matplotlib.pyplot as plt
plot_confusion_matrix(clf, X_test, y_test)
plt.xticks(rotation=90)
plt.tight_layout()
plt.show()