import os
import cv2
import numpy as np
from sklearn.decomposition import PCA
from sklearn.model_selection import train_test_split
from sklearn import svm
from sklearn.preprocessing import StandardScaler
import matplotlib.pyplot as plt

FEATURE_VECTOR_SIZE = 25344 # 144x176

dneg_folder_path = "/home/eric/URA/OASIS/Dataset/BMP/DNEG/"
dpos_folder_path = "/home/eric/URA/OASIS/Dataset/BMP/DPOS/"

dneg_files = []
dpos_files = []

for file in os.listdir(dneg_folder_path):
    dneg_file_path = os.path.join(dneg_folder_path, file)
    if os.path.isfile(dneg_file_path):
        dneg_img = cv2.imread(dneg_file_path, 0)
        dneg_files.append(dneg_img)

for file in os.listdir(dpos_folder_path):
    dpos_file_path = os.path.join(dpos_folder_path, file)
    if os.path.isfile(dpos_file_path):
        dpos_img = cv2.imread(dpos_file_path, 0)
        dpos_files.append(dpos_img)

# Create labels
dneg_labeled_files = [0] * len(dneg_files)
dpos_labeled_files = [1] * len(dpos_files)

# Concatenate arrays
data = dneg_files + dpos_files
labels = dneg_labeled_files + dpos_labeled_files

# Convert arrays to numpy arrays
data = np.array(data)
labels = np.array(labels)

# Flatten and scale data
data_flatten = data.reshape(len(data), FEATURE_VECTOR_SIZE)
scaler = StandardScaler()
data_flatten = scaler.fit_transform(data_flatten)

# Reduce dimensions
pca = PCA(n_components=115)
pca.fit(data_flatten)
data_flatten = pca.transform(data_flatten)
# print(data_flatten)

# PC_values = np.arange(pca.n_components_) + 1
# plt.plot(PC_values, pca.explained_variance_ratio_, 'o-', linewidth=2, color='blue')
# plt.title('Scree Plot')
# plt.xlabel('Principal Component')
# plt.ylabel('Variance Explained')
# plt.show()

for x in range(50):
	# Split data into training and testing sets
	data_training, data_testing, labels_training, labels_testing = train_test_split(data_flatten, labels, test_size = 0.2)

	# Train and evaluate SVM
	clf = svm.SVC()
	clf.fit(data_training, labels_training)
	print("C value: ", clf.get_params()['C'])
	print("Gamma value: ", clf.get_params()['gamma'])
	print(clf.score(data_testing, labels_testing))

	# Plot data
	# plt.scatter(data_flatten[:, 0], data_flatten[:, 1], c=labels, cmap='viridis')
	# plt.show()

	# accuracies = []
	# for num_components in range(2, 234):
	#     # Flatten and scale data
	#     data_flatten = data.reshape(len(data), FEATURE_VECTOR_SIZE)
	#     scaler = StandardScaler()
	#     data_flatten = scaler.fit_transform(data_flatten)

	#     # Reduce dimensions
	#     pca = PCA(n_components=num_components)
	#     pca.fit(data_flatten)
	#     data_flatten_pca = pca.transform(data_flatten)

	#     # Calculate accuracy 10 times
	#     accuracy = 0
	#     for i in range(100):
	#         # Split data into training and testing sets
	#         data_training, data_testing, labels_training, labels_testing = train_test_split(data_flatten_pca, labels, test_size = 0.2)

	#         # Train and evaluate SVM
	#         clf = svm.SVC()
	#         clf.fit(data_training, labels_training)
	#         accuracy += clf.score(data_testing, labels_testing)
	#     accuracy /= 100
	#     accuracies.append(accuracy)

	# # Plot results
	# plt.plot(range(2, 234), accuracies)
	# plt.title('Number of PCA Components vs. SVM Accuracy')
	# plt.xlabel('Number of PCA Components')
	# plt.ylabel('Accuracy')
	# plt.show()