def train(alpha, n_iter, plot=False):
size_img, rows, cols, images = utils.read_images(
TRAINING_IMAGES_PATH
)
size_lbl, labels = utils.read_labels(TRAINING_LABELS_PATH)
images = images[range(0, M_TRAINING), :]
labels = labels[range(0, M_TRAINING), :]
size_img = M_TRAINING
bias_terms = np.ones([size_img, 1], dtype=np.float64)
images = np.concatenate((bias_terms, images), axis=1).astype(np.float64)
thetas = np.zeros([rows*cols+1, N_LABELS], dtype=np.float64)
costs = np.zeros([n_iter, N_LABELS])
X = images / 255
for i in range(N_LABELS):
# print 'Training a classifier for label {0}'.format(i)
y = np.array([[1 if label == i else 0 for label in labels]]).T
thetas[:, i:i+1], costs[:, i:i+1] = func.gradient_descent(
thetas[:, i:i+1],
y, X, alpha,
n_iter
)
if plot:
plt.plot(costs[:, i:i+1])
plt.show()
return thetas
def test(thetas):
size_img, rows, cols, images = utils.read_images(
TRAINING_IMAGES_PATH
)
size_lbl, labels = utils.read_labels(TRAINING_LABELS_PATH)
bias_terms = np.ones([size_img, 1])
images = np.concatenate((bias_terms, images), axis=1)
training_images = images[range(0, M_TRAINING), :]
training_labels = labels[range(0, M_TRAINING), :]
X = training_images / 255
y = training_labels
accuracy = _test(thetas, y, X, M_TRAINING)
print 'Trained with {0} examples. Training accuracy: {1}'.format(
M_TRAINING,
accuracy
)
test_labels = labels[range(M_TRAINING, M_TRAINING + M_TESTING), :]
test_images = images[range(M_TRAINING, M_TRAINING + M_TESTING), :]
X = test_images / 255
y = test_labels
accuracy = _test(thetas, y, X, M_TESTING)
print 'Tested with {0} examples. Test accuracy: {1}'.format(M_TESTING,
accuracy)
def test():
x = tf.placeholder(tf.float32, shape=[1, 32, 32, 1], name='input')
network = get_network(args.network)(x=x, is_training=False)
testing_images = os.path.join(args.testing_folder, 'images')
testing_labels = read_labels(
os.path.join(args.testing_folder, 'labels.txt'))
loader = tf.train.Saver()
config = tf.ConfigProto(allow_soft_placement=True)
correct_predictions = 0
total_number = len(testing_labels)
with tf.Session(config=config) as session:
loader.restore(session, args.model)
for step in range(total_number):
next_batch = get_test_batch(testing_images, testing_labels, step)
predicted_value = session.run(network.prediction,
feed_dict={x: next_batch['images']})
if np.squeeze(predicted_value) == testing_labels[step]:
correct_predictions += 1
print(
'Testing results: {}/{} correct predictions Error rate:{}'.format(
correct_predictions, total_number,
100. * (total_number - correct_predictions) / total_number))
def calc_fsaverage_labels_indices(surf_name='pial', labels_from_annot=False, labels_fol='', parc='aparc250', subjects_dir=''):
labels_fol = os.path.join(subjects_dir, 'fsaverage', 'label', parc) if labels_fol=='' else labels_fol
if (labels_from_annot):
labels = mne.read_labels_from_annot('fsaverage', parc=parc, hemi='both', surf_name=surf_name)
else:
labels = utils.read_labels(labels_fol)
fsave_vertices = [np.arange(10242), np.arange(10242)]
labels_vertices, labels_names = utils.get_labels_vertices(labels, fsave_vertices)
np.savez(op.join(LOCAL_ROOT_DIR, 'fsaverage_labels_indices'), labels_vertices=labels_vertices, labels_names=labels_names)
def __getitem__(self, item):
image = Image.open(self.image_list[item]).convert('RGB')
image = self.transform(image)
label = utils.read_labels(self.label_list[item])
# convert to S*S*5 Tensor with format <x> <y> <w> <h> <cls>
label = utils.labels2tensor(label)
# get filename
filename = self.image_list[item].split('/')[-1]
return image, label, filename
def __init__(self, model_path, label_path, score_threshold):
"""
Initiates the ObjectDetectionEngine class.
:param model_path: Path to SSD object detection model
:param label_path: Path to coco format labels
:param score_threshold: Threshold for detection in the form of a float between 0 and 1
"""
InferenceEngine.__init__(self, model_path, [(1, 10, 4), (1, 10),
(1, 10), (1)])
self.labels = read_labels(label_path)
self.score_threshold = score_threshold
def run_evaluation(test_labels, predicted_labels):
"""
Gives statistical information on the performance of given algorithm.
:param list test_labels: Real value of test labels
:param str predicted_labels: Predicted value of test labels
:return: Confusion Matrix, Accuracy, Precision, Recall, fMeasure
"""
predicted_labels = utils.read_labels(predicted_labels)
co_matrix = confusion_matrix(test_labels, predicted_labels)
accuracy = utils.estimate_accuracy(test_labels, predicted_labels)
precision_recall_fscore = precision_recall_fscore_support(
test_labels, predicted_labels, average='weighted')
print(class_predictions(test_labels, predicted_labels))
return co_matrix, accuracy, precision_recall_fscore
def load_workspace(self, path):
"""
Sets the workspace directory. Checks for missing directories and files,
then loads all images, annotations & label list.
:param path: The path of the directory
"""
if not path:
path = "/tmp"
self.workspace = path
self.output_path_edit.setText(path)
# clear all lists and references
self.labels = []
self.cur_annotation_index = -1
self.cur_annotated_image = None
self.images_with_annotations = []
self.changes_done = False
self.class_id = -1
self.label = ""
# clear combo box and listWidgets
self.option_selector.clear()
self.image_list_widget.clear()
self.annotation_list_widget.clear()
if utils.check_workspace(self.workspace):
image_dir = self.workspace + "/images"
label_file = self.workspace + "/labels.txt"
self.labels = utils.read_labels(label_file)
for label in self.labels:
self.option_selector.addItem(label[0])
for dirname, dirnames, filenames in os.walk(image_dir):
for filename in sorted(filenames):
image_file = dirname + '/' + filename
label_file = image_file.replace(
"images",
"labels").replace(".jpg",
".txt").replace(".png", ".txt")
annotated_image = utils.read_annotated_image(
image_file, label_file)
self.images_with_annotations.append(annotated_image)
self.add_image_to_list_widget("/images/" + filename)
print("workspace loaded successfully")
import xml.etree.ElementTree as et
import numpy as np
import tensorflow as tf
import time
import sys
import pickle
import config
import utils
utils.read_labels("yelp")
import class_DatasetAgN as ds
import cnn as cn
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.decomposition import TruncatedSVD
from sklearn.pipeline import Pipeline
from stop_words import get_stop_words
from sklearn.metrics import pairwise_distances
env = sys.argv[1]
config.dictionary_size = 1014
config.vocabulary_size = 1
# print(labels)
print("Total labels: ", len(config.labels))
print(config.vocabulary_size)
path = ""
if env == "local":
path = "data/reuters/"
elif env == "server":
path = "data/reuters/"
TEST_DIR = "test/"
# use HOG as a list of features
# reading in the data. This takes a while
train_imgs = utils.read_folder(TRAIN_DIR, 0, ntrain, flatten = False)
print ("\nDone!")
sys.stdout.flush()
print ("Getting HOG3 of the data...")
sys.stdout.flush()
# also takes a while
X = utils.getHOG3(train_imgs)
print ("\nDone!")
sys.stdout.flush()
X = np.insert(X, 0, 1.0, axis = 1)
theta = np.random.randn(X.shape[1], 10) * 0.0001
y = utils.read_labels('trainLabels.csv', 0, ntrain)
best_val = -1
best_softmax = None
X_train, X_val, y_train, y_val = cross_validation.train_test_split(X, y, test_size = 0.1)
print "y_train.shape=", y_train.shape
print "y_val.shape=", y_val.shape
print "X_train.shape=", X_train.shape
print "X_val.shape=", X_val.shape
sys.stdout.flush()
# this part was used to the best hyperparameters
#best LR = [15]
learning_rates = [15] # [11,12,13,14,15,16,17,18,19,20]
#Best RS = [1e-3]
from model import LeNet
parser = argparse.ArgumentParser(
description='Training LeNet on MNIST using NumPy.')
parser.add_argument('data_dir', type=str, help='directory to mnist data')
args = parser.parse_args()
epoch = 10
lr = 0.1
momentum = 0.8
batch = 256
train_data = read_images(os.path.join(args.data_dir,
'train-images.idx3-ubyte'))
train_labels = read_labels(
os.path.join(args.data_dir, 'train-labels.idx1-ubyte'))
test_data = read_images(os.path.join(args.data_dir, 't10k-images.idx3-ubyte'))
test_labels = read_labels(os.path.join(args.data_dir,
't10k-labels.idx1-ubyte'))
# normalize
train_data = (train_data - train_data.mean(
(1, 2), keepdims=True)) / train_data.std((1, 2), keepdims=True)
test_data = (test_data - test_data.mean(
(1, 2), keepdims=True)) / test_data.std((1, 2), keepdims=True)
my_net = LeNet()
optimizer = SGD(my_net.parameters(), lr, momentum)
loss_history = []
def main():
""" Main function
"""
parser = ArgumentParser(
description='Affective Computing with AMIGOS Dataset -- XGBoost')
parser.add_argument('--data', type=str, default='./data')
parser.add_argument('--feat',
type=str,
choices=['eeg', 'ecg', 'gsr', 'all'],
default='all',
help='choose type of modality')
parser.add_argument('--old', action='store_true')
args = parser.parse_args()
# read extracted features
if args.old:
data = np.loadtxt(os.path.join(args.data, 'features.csv'),
delimiter=',')[:, :213]
if args.feat == 'eeg':
data = data[:, 108:213]
elif args.feat == 'ecg':
data = data[:, 32:108]
elif args.feat == 'gsr':
data = data[:, 0:32]
a_data, v_data = (data, data)
else:
a_data = np.loadtxt(os.path.join(args.data, 'a_features.csv'),
delimiter=',')
v_data = np.loadtxt(os.path.join(args.data, 'v_features.csv'),
delimiter=',')
if args.feat == 'eeg':
a_data = np.hstack((a_data[:, 108:213], a_data[:, 229:232]))
v_data = np.hstack((v_data[:, 108:213], v_data[:, 217:316]))
elif args.feat == 'ecg':
a_data = np.hstack(
(a_data[:, 32:108], a_data[:, 213:215], a_data[:, 232:234]))
v_data = np.hstack(
(v_data[:, 32:108], v_data[:, 213:217], v_data[:, 316:324]))
elif args.feat == 'gsr':
a_data = np.hstack(
(a_data[:, 0:32], a_data[:, 215:229], a_data[:, 234:254]))
v_data = v_data[:, 0:32]
# read labels and split to 0 and 1 by
a_labels, v_labels = read_labels(os.path.join(args.data, 'label.csv'))
# setup kfold cross validator
sub_num = SUBJECT_NUM - len(MISSING_DATA_SUBJECT)
kf = KFold(n_splits=sub_num)
# classifier parameters
with open(
os.path.join(
args.data, 'model',
"{}_a_{}_model.pkl".format('old' if args.old else 'new',
args.feat)), 'rb') as f:
a_params = pickle.load(f)
with open(
os.path.join(
args.data, 'model',
"{}_v_{}_model.pkl".format('old' if args.old else 'new',
args.feat)), 'rb') as f:
v_params = pickle.load(f)
a_clf = xgb.XGBClassifier(objective="binary:logistic")
v_clf = xgb.XGBClassifier(objective="binary:logistic")
a_clf.set_params(**a_params)
v_clf.set_params(**v_params)
train_a_accuracy_history = []
train_v_accuracy_history = []
val_a_accuracy_history = []
val_v_accuracy_history = []
for fold, (train_idx, val_idx) in enumerate(kf.split(a_data)):
print('Fold', fold + 1)
# normalize using mean and std
a_data = (a_data - np.mean(a_data, axis=0)) / np.std(a_data, axis=0)
v_data = (v_data - np.mean(v_data, axis=0)) / np.std(v_data, axis=0)
# collect data for cross validation
a_train_data, a_val_data = a_data[train_idx], a_data[val_idx]
v_train_data, v_val_data = v_data[train_idx], v_data[val_idx]
train_a_labels, val_a_labels = a_labels[train_idx], a_labels[val_idx]
train_v_labels, val_v_labels = v_labels[train_idx], v_labels[val_idx]
# fit classifier
a_clf.fit(a_train_data, train_a_labels)
v_clf.fit(v_train_data, train_v_labels)
# predict arousal and valence
train_a_predict_labels = a_clf.predict(a_train_data)
train_v_predict_labels = v_clf.predict(v_train_data)
val_a_predict_labels = a_clf.predict(a_val_data)
val_v_predict_labels = v_clf.predict(v_val_data)
# metrics calculation
train_a_accuracy = f1_score(train_a_labels,
train_a_predict_labels,
average='macro')
train_v_accuracy = f1_score(train_v_labels,
train_v_predict_labels,
average='macro')
val_a_accuracy = f1_score(val_a_labels,
val_a_predict_labels,
average='macro')
val_v_accuracy = f1_score(val_v_labels,
val_v_predict_labels,
average='macro')
train_a_accuracy_history.append(train_a_accuracy)
train_v_accuracy_history.append(train_v_accuracy)
val_a_accuracy_history.append(val_a_accuracy)
val_v_accuracy_history.append(val_v_accuracy)
print('Training Result:')
print("Arousal: acc = {:.4f}".format(np.mean(train_a_accuracy_history)))
print("Valence: acc = {:.4f}\n".format(np.mean(train_v_accuracy_history)))
print('Validating Result:')
print("Arousal: acc = {:.4f}".format(np.mean(val_a_accuracy_history)))
print("Valence: acc = {:.4f}\n".format(np.mean(val_v_accuracy_history)))
a_imp = a_clf.feature_importances_
v_imp = v_clf.feature_importances_
a_sort_idx = np.argsort(a_imp)[::-1]
v_sort_idx = np.argsort(v_imp)[::-1]
with open('a_imp_name', 'w') as f:
for idx in a_sort_idx:
if a_imp[idx] != 0:
f.write("{}\n".format(A_FEATURE_NAMES[idx]))
with open('v_imp_name', 'w') as f:
for idx in v_sort_idx:
if v_imp[idx] != 0:
f.write("{}\n".format(V_FEATURE_NAMES[idx]))
import random
import network
import utils
labels = utils.read_labels()
utils.prepare_images()
all_images = utils.get_images()
image = utils.load_image(all_images[0])
input_shape = image.shape
BATCH_SIZE = 32
EPOCHS = 50
def load_pair(fname):
f = utils.load_image(fname + '_firefox.png')
print(f.shape)
c = utils.load_image(fname + '_chrome.png')
print(c.shape)
return [f, c]
images_train = random.sample(all_images, int(len(all_images) * 0.9))
images_test = [i for i in all_images if i not in set(images_train)]
def couples_generator(images):
for i in images:
yield load_pair(i), labels[i]
import argparse
from PIL import ImageTk, Image
from tkinter import Tk, Label
import utils
labels_directory = "label_persons/"
parser = argparse.ArgumentParser()
parser.add_argument("file_name", action="store")
args = parser.parse_args()
labels = utils.read_labels(labels_directory + args.file_name)
images_to_show = [i for i in utils.get_images() if i not in labels]
current_image = None
root = Tk()
panel1 = Label(root)
panel1.pack(side="left", padx=10)
panel2 = Label(root)
panel2.pack(side="left", padx=10)
def get_new_image():
global current_image
if len(images_to_show) == 0:
root.quit()
return
current_image = images_to_show.pop()
print("data/%s_firefox.png" % current_image)
# mean from dataset058:
mean = [0.3823625683879477, 0.3790166856065496, 0.3554138533338805]
std=[0.21754145353302254, 0.21271749678359336, 0.21233947166469555]
#mean = [0.383661700858527, 0.3819784115384924, 0.3588786631614881]
#std=[0.2167717755518767, 0.21201058526724945, 0.21143164036556178]
# Load the dataset
from utils import read_images_stl10 as read_images
from utils import read_labels_stl10 as read_labels
#unlab_set_x = read_images('../data/stl10_binary/unlabeled_X.bin')
test_set_x = read_images('../../data/stl10_binary/test_X.bin')
train_set_x = read_images('../../data/stl10_binary/train_X.bin')
test_set_y = read_labels('../../data/stl10_binary/test_y.bin')
train_set_y = read_labels('../../data/stl10_binary/train_y.bin')
print 'Train set information: '
print (len(train_set_x), type(train_set_x[3]))
print ''
print 'Test set information: '
print (len(test_set_x), type(test_set_x[3]))
print ''
# loading the folds:
with open('../../data/stl10_binary/fold_indices.txt', 'r') as f_folds:
folds = f_folds.readlines()
k_folds = []
for fold_i in folds:
import xml.etree.ElementTree as et
import numpy as np
import tensorflow as tf
import time
import sys
import pickle
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfVectorizer
from stop_words import get_stop_words
import config
import utils
config.max_characters = 1932
config.vocabulary_size = 4
utils.read_labels("rcv")
import class_DatasetRcv as ds
import cnn as cn
if len(sys.argv) > 1:
env = sys.argv[1]
else:
env = "local"
# print(labels)
print("Total labels: ", len(config.labels))
print(config.vocabulary_size)
path = ""
if env == "local":
path = "data/reuters/"
elif env == "server":
path = "data/reuters/"
def train():
global_step = tf.Variable(0, trainable=False)
x = tf.placeholder(tf.float32, shape=[args.batch, 32, 32, 1], name='input')
labels = tf.placeholder(tf.float32, shape=[args.batch, 10], name='labels')
network = get_network(args.network)(x=x, labels=labels)
optimizer = tf.train.AdamOptimizer(learning_rate=args.lr)
loss = network.loss
optimization_op = optimizer.minimize(loss)
saver = tf.train.Saver()
number_of_params = 0
for variable in tf.trainable_variables():
number_of_params += np.array(variable.get_shape().as_list()).prod()
print("number of trainable parameters: {}".format(number_of_params))
training_images = os.path.join(args.training_folder, 'images')
training_labels = read_labels(
os.path.join(args.training_folder, 'labels.txt'))
validation_images = os.path.join(args.validation_folder, 'images')
validation_labels = read_labels(
os.path.join(args.validation_folder, 'labels.txt'))
config = tf.ConfigProto(allow_soft_placement=True)
start = time.time()
best_accuracy = 0
with tf.Session(config=config) as session:
session.run(tf.global_variables_initializer())
session.run(tf.local_variables_initializer())
valid_images = []
for step in range(args.step):
next_batch = get_training_batch(training_images,
training_labels,
batch_size=args.batch)
loss_value, _ = session.run([loss, optimization_op],
feed_dict={
x: next_batch['images'],
labels: next_batch['labels']
})
elapsed_time = time.time() - start
if step and step % 1000 == 0:
print(
'step:{}/{} | loss:{:7.6f} | elapsed time: {:2.0f}m:{:2.0f}s'
.format(step, args.step, loss_value, elapsed_time // (60),
int(elapsed_time) % 60))
if step and args.validation_step > 0 and step % args.validation_step == 0:
good_predictions = 0
for validation_step in range(0, len(validation_labels),
args.batch):
next_validation_batch = get_validation_batch(
validation_images,
validation_labels,
batch_size=args.batch,
starting_index=validation_step)
predicted_value = session.run(
network.prediction,
feed_dict={
x: next_validation_batch['images'],
labels: next_validation_batch['labels']
})
gts = np.argmax(next_validation_batch['labels'], axis=-1)
good_predictions += (predicted_value == gts).sum()
print('Validation: {}/{} good predictions'.format(
good_predictions, len(validation_labels)))
if good_predictions >= best_accuracy:
saver.save(session, './trained_models/best')
best_accuracy = good_predictions
saver.save(session, './trained_models/model', global_step=args.step)
print('Training ended!')
@app.route('/download_report')
def download_report():
try:
return send_file("/home/pi/tflite1/bir_isim/no_tracking/report/" +
report_file_name)
except FileNotFoundError:
abort(404)
@app.route('/')
def index():
return render_template('index.html')
labels = read_labels(PATH_TO_LABELS)
interpreter = initialize_detector(MODEL_NAME)
def m():
global output, heatmap_image, total_people, field1_count, field2_count
frame_rate_calc = 1
freq = cv2.getTickFrequency()
videostream = VideoStream(VIDEO_PATH).start()
color_f1 = (0, 0, 255)
color_f2 = (255, 0, 0)
heatmap = np.zeros((720, 1270, 3), dtype=np.uint8)
ht_color = (191, 255, 1)
# cols = original_score[:,1].astype(int)
# computed_scores = A[rows, cols]
# D = np.eye(len(A))
# D[rows, cols] = computed_scores
# D[cols, rows] = computed_scores
# return D
#############################################################################################
#================================= Data extraction =========================================#
# READ THE FILE
dataset = sys.argv[1]
num_samples = int(sys.argv[3])
reshaped_preds = read_file(file_="../GYPSUM/" + dataset + "_predicts_0.npy")
# READ LABELS FROM FILE
original_scores = read_labels("../GYPSUM/" + dataset + "_label_ids.txt")
# reshaped_preds = reshaped_preds
# reshaped_preds = fix_vals_first(reshaped_preds, original_scores)
# print(reshaped_preds.shape)
# FIND OUT DUPLICATE ROWS AND ELIMINATE!!
unique_rows, unique_indices = np.unique(reshaped_preds,
axis=0,
return_index=True)
# FIND DUPLICATE ROWS AND ID THEM WITH MATCHING ROWS
# Duplicate ids == rows in reshaped_preds which has been removed
# Matched ids = rows in reshaped_preds which has a duplicate in duplcate_ids
unique_indices = np.sort(unique_indices)
# checked, and the following line works fine
import numpy as np
import tensorflow as tf
import time
import pickle
import sys
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfVectorizer
import config
import utils
utils.read_labels("bibtex")
import class_DatasetRcv as ds
import encoder as au
import mlpau as ml
from stop_words import get_stop_words
from tensorflow.python.framework import ops
from tensorflow.python.ops import array_ops
from tensorflow.python.ops import sparse_ops
env = sys.argv[1]
bpmll_out_module = tf.load_op_library('custom/bp_mll.so')
bpmll_grad_out_module = tf.load_op_library('custom/bp_mll_grad.so')
@ops.RegisterGradient("BpMll")
def _bp_mll_grad(op, grad):
return bpmll_grad_out_module.bp_mll_grad(grad=grad,
logits=op.inputs[0],
labels=op.inputs[1])
def main():
""" Main function """
# read extracted features
amigos_data = np.loadtxt('data/mde_features.csv', delimiter=',')
eeg_entropy = amigos_data[:, :200]
ecg_entropy = amigos_data[:, 200:206]
gsr_entropy = amigos_data[:, 206:]
# read labels and split to 0 and 1 by local mean
a_labels, v_labels = read_labels('data/signals/label.csv')
pos_a_idx = np.where(a_labels == 1)[0]
neg_a_idx = np.where(a_labels == 0)[0]
pos_v_idx = np.where(v_labels == 1)[0]
neg_v_idx = np.where(v_labels == 0)[0]
# EEG
for d in range(2, 4):
for r in range(1, 6):
for typ in ['a', 'v']:
start_idx = (d - 2) * 5 + (r - 1)
end_idx = 20 * 2 * 5
if typ == 'a':
pos_values = eeg_entropy[pos_a_idx, start_idx:end_idx:10]
neg_values = eeg_entropy[neg_a_idx, start_idx:end_idx:10]
elif typ == 'v':
pos_values = eeg_entropy[pos_v_idx, start_idx:end_idx:10]
neg_values = eeg_entropy[neg_v_idx, start_idx:end_idx:10]
_, ax = plt.subplots()
for axis in [ax.xaxis, ax.yaxis]:
axis.set_major_locator(ticker.MaxNLocator(integer=True))
ax.errorbar(np.arange(1, 21),
np.mean(pos_values, axis=0) / NORM,
yerr=np.std(pos_values) / pos_values.shape[0],
marker='o',
markersize=4,
label='pos',
color='#E76F51',
capsize=4)
ax.errorbar(np.arange(1, 21),
np.mean(neg_values, axis=0) / NORM,
yerr=np.std(neg_values) / neg_values.shape[0],
marker='o',
markersize=4,
label='neg',
color='#44A261',
capsize=4)
ax.legend()
ax.margins(0.05)
ax.axis('tight')
plt.tight_layout()
plt.savefig("plot/mde/eeg_mmde/{}_eeg_d{}_r{}.png".format(
typ, d, r))
plt.close()
# ECG
for d in range(2, 4):
for typ in ['a', 'v']:
start_idx = d - 2
end_idx = 3 * 2
if typ == 'a':
pos_values = ecg_entropy[pos_a_idx, start_idx:end_idx:2]
neg_values = ecg_entropy[neg_a_idx, start_idx:end_idx:2]
elif typ == 'v':
pos_values = ecg_entropy[pos_v_idx, start_idx:end_idx:2]
neg_values = ecg_entropy[neg_v_idx, start_idx:end_idx:2]
_, ax = plt.subplots()
for axis in [ax.xaxis, ax.yaxis]:
axis.set_major_locator(ticker.MaxNLocator(integer=True))
ax.errorbar(np.arange(1, 4),
np.mean(pos_values, axis=0) / NORM,
yerr=np.std(pos_values) / pos_values.shape[0],
marker='o',
markersize=4,
label='pos',
color='#E76F51',
capsize=4)
ax.errorbar(np.arange(1, 4),
np.mean(neg_values, axis=0) / NORM,
yerr=np.std(neg_values) / neg_values.shape[0],
marker='o',
markersize=4,
label='neg',
color='#44A261',
capsize=4)
ax.legend()
ax.margins(0.05)
ax.axis('tight')
plt.tight_layout()
plt.savefig("plot/mde/ecg_rcmde/{}_ecg_d{}.png".format(typ, d))
plt.close()
# GSR
for d in range(2, 4):
for typ in ['a', 'v']:
start_idx = d - 2
end_idx = 20 * 2
if typ == 'a':
pos_values = gsr_entropy[pos_a_idx, start_idx:end_idx:2]
neg_values = gsr_entropy[neg_a_idx, start_idx:end_idx:2]
elif typ == 'v':
pos_values = gsr_entropy[pos_v_idx, start_idx:end_idx:2]
neg_values = gsr_entropy[neg_v_idx, start_idx:end_idx:2]
_, ax = plt.subplots()
for axis in [ax.xaxis, ax.yaxis]:
axis.set_major_locator(ticker.MaxNLocator(integer=True))
ax.errorbar(np.arange(1, 21),
np.mean(pos_values, axis=0) / NORM,
yerr=np.std(pos_values) / pos_values.shape[0],
marker='o',
markersize=4,
label='pos',
color='#E76F51',
capsize=4)
ax.errorbar(np.arange(1, 21),
np.mean(neg_values, axis=0) / NORM,
yerr=np.std(neg_values) / neg_values.shape[0],
marker='o',
markersize=4,
label='neg',
color='#44A261',
capsize=4)
ax.legend()
ax.margins(0.05)
ax.axis('tight')
plt.tight_layout()
plt.savefig("plot/mde/gsr_rcmde/{}_gsr_d{}.png".format(typ, d))
plt.close()
from index import run_script, run_knn, run_logistic, run_naive_bayes, run_tree, run_evaluation
from utils import read_labels, show_estimation_utility
import constants
data1 = input('Enter data-set path for training. (Ex: file.txt): ')
data2 = input('Enter data-set path for training: ')
chosen_model = input(
'Enter model name for testing. '
'Options: KNN, naiveBayes, logisticRegression, treeClassifier, finalModel: '
)
test_file = input('Enter test file: ')
test_labels_file = input('Enter test labels file: ')
tweets, data_vectors, test_vector = run_script(data1, data2, test_file)
test_labels = read_labels(test_labels_file)
if chosen_model == 'KNN':
run_knn(data_vectors, test_vector)
confusion_matrix, accuracy, precision_recall_fscore = run_evaluation(
test_labels, constants.KNN_PREDICTION)
show_estimation_utility(confusion_matrix, accuracy,
precision_recall_fscore)
elif chosen_model == 'naiveBayes':
run_naive_bayes(test_vector, constants.BAYES_MODEL)
confusion_matrix, accuracy, precision_recall_fscore = run_evaluation(
test_labels, constants.BAYES_PREDICTION)
show_estimation_utility(confusion_matrix, accuracy,
precision_recall_fscore)
elif chosen_model == 'logisticRegression':
run_logistic(test_vector)
confusion_matrix, accuracy, precision_recall_fscore = run_evaluation(
print(imgList[:5])
for img in imgList:
im = cv2.imread(img, cv2.IMREAD_COLOR).astype(np.float32) / 255.
im = np.reshape(
im, [FLAGS.image_height, FLAGS.image_width, FLAGS.image_channel])
imgs_input.append(im)
imgs_input = np.asarray(imgs_input)
model = cnn_lstm_otc_ocr.LSTMOCR('infer')
model.build_graph()
# 定义节点
logit = model.get_logist()
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver(tf.global_variables(), max_to_keep=100)
ckpt = tf.train.latest_checkpoint(FLAGS.checkpoint_dir)
saver.restore(sess, ckpt)
print('restore from ckpt {}'.format(ckpt))
feed = {model.inputs: imgs_input}
dense_decoded_code = sess.run(model.dense_decoded, feed)
decoded_expression = utils.read_labels(dense_decoded_code)
print(sess.run(logit, feed))
for fn, code in zip(imgList, decoded_expression):
print(fn + ' : ' + code)
def main():
""" Main function
"""
parser = ArgumentParser(
description='Affective Computing with AMIGOS Dataset -- Tuning')
parser.add_argument('--data', type=str, default='./data')
parser.add_argument('--feat',
type=str,
choices=['eeg', 'ecg', 'gsr', 'all'],
default='all',
help='choose type of modality')
parser.add_argument('--clf',
type=str,
choices=['gnb', 'svm', 'xgb'],
default='xgb',
help='choose type of classifier')
parser.add_argument('--old', action='store_true')
args = parser.parse_args()
# read extracted features
if args.old:
data = np.loadtxt(os.path.join(args.data, 'features.csv'),
delimiter=',')[:, :213]
if args.feat == 'eeg':
data = data[:, 108:213]
elif args.feat == 'ecg':
data = data[:, 32:108]
elif args.feat == 'gsr':
data = data[:, 0:32]
amigos_data = (data, data)
else:
a_data = np.loadtxt(os.path.join(args.data, 'a_features.csv'),
delimiter=',')
v_data = np.loadtxt(os.path.join(args.data, 'v_features.csv'),
delimiter=',')
if args.feat == 'eeg':
a_data = np.hstack((a_data[:, 108:213], a_data[:, 229:232]))
v_data = np.hstack((v_data[:, 108:213], v_data[:, 217:316]))
elif args.feat == 'ecg':
a_data = np.hstack(
(a_data[:, 32:108], a_data[:, 213:215], a_data[:, 232:234]))
v_data = np.hstack(
(v_data[:, 32:108], v_data[:, 213:217], v_data[:, 316:324]))
elif args.feat == 'gsr':
a_data = np.hstack(
(a_data[:, 0:32], a_data[:, 215:229], a_data[:, 234:254]))
v_data = v_data[:, 0:32]
amigos_data = (a_data, v_data)
# read labels and split to 0 and 1 by
a_labels, v_labels = read_labels(os.path.join(args.data, 'label.csv'))
labels = (a_labels, v_labels)
# setup kfold cross validator
sub_num = SUBJECT_NUM - len(MISSING_DATA_SUBJECT)
kf = KFold(n_splits=sub_num)
# tune classifier parameters
grid_search_params = {'max_depth': [], 'n_estimators': []}
other_tuning_params = {
'learning_rate': np.arange(0.01, 0.41, 0.01),
'gamma': np.arange(0, 10.1, 0.5),
'min_child_weight': np.arange(0.80, 1.21, 0.01),
'max_delta_step': np.arange(0, 2.05, 0.05),
'subsample': np.arange(1.00, 0.59, -0.01),
'colsample_bytree': np.arange(1.00, 0.09, -0.01),
'colsample_bylevel': np.arange(1.00, 0.09, -0.01),
'reg_alpha': np.arange(0, 2.05, 0.05),
'reg_lambda': np.arange(0.50, 2.55, 0.05),
'scale_pos_weight': np.arange(0.80, 1.21, 0.01),
'base_score': np.arange(0.40, 0.61, 0.01),
'seed': np.arange(0, 41)
}
# with open(os.path.join(args.data, 'model', "old_a_{}_model.pkl".format(args.feat)), 'rb') as f:
# a_param = pickle.load(f)
# with open(os.path.join(args.data, 'model', "old_v_{}_model.pkl".format(args.feat)), 'rb') as f:
# v_param = pickle.load(f)
# grid search tuning
a_best_params = {'max_depth': 3, 'n_estimators': 20}
v_best_params = {'max_depth': 3, 'n_estimators': 20}
a_acc, v_acc = 0, 0
print('Tuning max_depth and n_estimators')
for param in grid_search_params['max_depth']:
print('max_depth', param)
clf = {
'a': xgb.XGBClassifier(max_depth=param,
objective="binary:logistic"),
'v': xgb.XGBClassifier(max_depth=param,
objective="binary:logistic")
}
tuning_params = grid_search_params['n_estimators']
a_param, v_param, tmp_a_acc, tmp_v_acc = tuning(
clf, 'n_estimators', tuning_params, amigos_data, labels, kf)
if tmp_a_acc >= a_acc:
a_best_params['max_depth'] = param
a_best_params['n_estimators'] = a_param
a_acc = tmp_a_acc
if tmp_v_acc >= v_acc:
v_best_params['max_depth'] = param
v_best_params['n_estimators'] = v_param
v_acc = tmp_v_acc
# tune other parameters
for param_name, tuning_params in other_tuning_params.items():
print('Tuning', param_name)
clf = {
'a': xgb.XGBClassifier(objective="binary:logistic"),
'v': xgb.XGBClassifier(objective="binary:logistic")
}
clf['a'].set_params(**a_best_params)
clf['v'].set_params(**v_best_params)
a_param, v_param, _, _ = tuning(clf, param_name, tuning_params,
amigos_data, labels, kf)
a_best_params[param_name] = a_param
v_best_params[param_name] = v_param
ver = 'old' if args.old else 'new'
with open(
os.path.join(args.data, 'model',
"{}_a_{}_model.pkl".format(ver, args.feat)),
'wb') as f:
pickle.dump(a_best_params, f)
with open(
os.path.join(args.data, 'model',
"{}_v_{}_model.pkl".format(ver, args.feat)),
'wb') as f:
pickle.dump(v_best_params, f)
def __getitem__(
self, idx: int
) -> Tuple[torch.tensor, Dict[str, torch.tensor], str]:
"""
Get dataset item.
Parameters
----------
idx : int
Dataset item index
Returns
-------
Tuple[Tensor, Dict[str, Tensor], str]
(image, target, image_id)
"""
image_id = self.image_ids[idx]
image = cv2.imread(f"{self.image_dir}/{image_id}", cv2.IMREAD_COLOR)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB).astype(np.float32)
# test dataset must have some values so that transforms work.
target = {
"labels": torch.as_tensor([[0]], dtype=torch.float32),
"boxes": torch.as_tensor([[0, 0, 0, 0]], dtype=torch.float32),
}
# for train and valid test create target dict.
if self.mode != "test":
image_data = self.df.loc[self.df["image_id"] == image_id]
boxes = image_data[["x", "y", "x1", "y1"]].values
boxes = torch.as_tensor(boxes, dtype=torch.float32)
areas = image_data["area"].values
areas = torch.as_tensor(areas, dtype=torch.float32)
label_dict = read_labels(
to_absolute_path(self.cfg.data.labels_path)
)
labels = [
label_dict[label] for label in image_data["label"].values
]
labels = torch.as_tensor(labels, dtype=torch.int64)
iscrowd = torch.zeros((image_data.shape[0],), dtype=torch.int64)
target["boxes"] = boxes
target["labels"] = labels
target["image_id"] = torch.tensor([idx])
target["area"] = areas
target["iscrowd"] = iscrowd
if self.transforms:
image_dict = {
"image": image,
"bboxes": target["boxes"],
"labels": labels,
}
image_dict = self.transforms(**image_dict)
image = image_dict["image"]
target["boxes"] = torch.as_tensor(
image_dict["bboxes"], dtype=torch.float32
)
else:
image_dict = {
"image": image,
"bboxes": target["boxes"],
"labels": target["labels"],
}
image = self.transforms(**image_dict)["image"]
return image, target, image_id
text += utils.read_text(i)
utils.logger.info("text reading finished")
tokens = utils.tokenize_paragraph_w2v(text)
utils.logger.info("text tokenizing finished")
utils.compute_paragraph_word2vec(tokens,
vector_size=args.vector_size,
workers=multiprocessing.cpu_count(),
model_path=args.wm)
utils.logger.info("word2vec training finished")
elif args.train_cnn or args.predict_cnn or args.validate_cnn:
train_data = None
test_data = None
if args.train_cnn or args.validate_cnn:
loaded = False
train_labels = utils.read_labels(args.l)
utils.logger.info("labels reading finished")
if args.use_d2v and utils.utils.is_path_accessible(args.dd):
utils.logger.info(
"try to load from doc2vec train data from specified path")
train_data = utils.read_array(args.dd)
utils.logger.info("load doc2vec train data successfully")
loaded = True
elif args.use_w2v and utils.is_path_accessible(args.wd):
utils.logger.info(
"try to load from word2vec train data from specified path")
train_data = utils.read_array(args.wd)
utils.logger.info("load word2vec train data successfully")
loaded = True
from neuralnetwork import NeuralNetworkModel
import autograd.numpy as np
import pickle
import matplotlib.pyplot as plt
from utils import plt_image
TRAINING_IMAGES_PATH = 'mnistset/train-images-idx3-ubyte.gz'
TRAINING_LABELS_PATH = 'mnistset/train-labels-idx1-ubyte.gz'
TESTING_IMAGES_PATH = 'mnistset/t10k-images-idx3-ubyte.gz'
TESTING_LABELS_PATH = 'mnistset/t10k-labels-idx1-ubyte.gz'
MODELS_PATH = 'mnistmodels/'
print 'Reading images from %s' % TRAINING_IMAGES_PATH
images = utils.read_images(TRAINING_IMAGES_PATH)
print 'Reading labels from %s' % TRAINING_LABELS_PATH
labels = utils.read_labels(TRAINING_LABELS_PATH)
print 'Reading images from %s' % TESTING_IMAGES_PATH
testing_images = utils.read_images(TESTING_IMAGES_PATH)
print 'Reading labels from %s' % TESTING_LABELS_PATH
testing_labels = utils.read_labels(TESTING_LABELS_PATH)
training_images = images[5000:-5000]
training_labels = labels[5000:-5000]
validating_images = np.concatenate([images[:5000], images[-5000:]])
validating_labels = np.concatenate([labels[:5000], labels[-5000:]])
np.set_printoptions(suppress=True)
nnm = None
def create(layers=[28 * 28, 100, 10], batch_size=32, dropout=0.1):
print 'Creating neural network'
import predictor
import FGSM_utils
import utils
import sys
if __name__ == '__main__':
PGM_NUM = 154
pgms = [
utils.read_PGM(f"../data/pgm/{i}.pgm") for i in range(1, PGM_NUM + 1)
]
labels = utils.read_labels("../data/labels.txt", PGM_NUM)
params = predictor.read_params("../data/param.txt")
pred = predictor.Predictor(params)
EPS = float(sys.argv[1])
for i in range(PGM_NUM):
x, raw_x = pgms[i]
t = labels[i] - 1
pred.forward(x)
dx = pred.backward(t) # get the gradient vector
sign = FGSM_utils.sign(dx)
prx = FGSM_utils.perturb(raw_x, sign, EPS)
utils.write_PGM(f"../data/FGSM_{EPS:.3f}/{i+1}.pgm", prx)
from neuralnetwork import NeuralNetworkModel
import autograd.numpy as np
import pickle
import matplotlib.pyplot as plt
from utils import plt_image
TRAINING_IMAGES_PATH = 'mnistset/train-images-idx3-ubyte.gz'
TRAINING_LABELS_PATH = 'mnistset/train-labels-idx1-ubyte.gz'
TESTING_IMAGES_PATH = 'mnistset/t10k-images-idx3-ubyte.gz'
TESTING_LABELS_PATH = 'mnistset/t10k-labels-idx1-ubyte.gz'
MODELS_PATH = 'mnistmodels/'
print 'Reading images from %s' % TRAINING_IMAGES_PATH
images = utils.read_images(TRAINING_IMAGES_PATH)
print 'Reading labels from %s' % TRAINING_LABELS_PATH
labels = utils.read_labels(TRAINING_LABELS_PATH)
print 'Reading images from %s' % TESTING_IMAGES_PATH
testing_images = utils.read_images(TESTING_IMAGES_PATH)
print 'Reading labels from %s' % TESTING_LABELS_PATH
testing_labels = utils.read_labels(TESTING_LABELS_PATH)
training_images = images[5000:-5000]
training_labels = labels[5000:-5000]
validating_images = np.concatenate([images[:5000], images[-5000:]])
validating_labels = np.concatenate([labels[:5000], labels[-5000:]])
np.set_printoptions(suppress=True)
nnm = None
def create(layers=[28*28, 100, 10], batch_size=32, dropout=0.1):
print 'Creating neural network'
X = np.concatenate((self.descs[self.pos_candidates, :],
self.descs[self.neg_candidates]))
return Y, X
def get_n_negs(self, N_neg=None):
if (self._has_pos()) and (N_neg is None):
N_neg = self.xy_pos.shape[0]
elif (not self._has_pos()):
N_neg = self.default_N_negs
return N_neg
if __name__ == "__main__":
xy_pos = utls.read_labels('../dataset/labels/0.txt')
xy, descs = utls.read_feats('../feats/0.npz')
im = io.imread('../dataset/images/0.png')
sampler = Sampler(xy_pos, xy, descs)
Y, f, xy = sampler.sample()
import matplotlib.pyplot as plt
_, ax = plt.subplots()
ax.imshow(im, cmap='gray')
utls.plot_bboxes(xy[Y == 1, :], im.shape, ax, color='g')
utls.plot_bboxes(xy[Y == 0, :], im.shape, ax, color='r')
plt.show()
