def main():
# データセットの取得
datasets = get_datasets(DATASETS_CSV,
test_size=TEST_SIZE,
image_size=IMAGE_SIZE)
train_data, _, _, _ = datasets
# TODO
# 29 0.9961215257644653
# 30回目で過学習により5割減
# 30回に分けてトレーニング実行
max_steps = 25
run_training(
datasets,
tensorboard_path=TENSORBOARD_PATH,
checkpoint_path=MODEL_PATH,
# 分類数
num_classes=len(get_labels(DATASETS_CSV)),
# 画像サイズ
image_size=IMAGE_SIZE,
# ピクセルのベクトル数 3=カラー,1=モノクロ
channel=CHANNELS,
# 学習実行回数
max_steps=max_steps,
# 1エポックで学習するデータサイズ
batch_size=int(len(train_data) / max_steps),
# 学習率
learning_rate=1e-4)
def mkdir(path):
folder = os.path.exists(path)
if not folder:
os.mkdir(path)
print("====new folder====")
print("========ok========")
else:
print("----There is this folder! ----")
mkdir("pic_save")
detection_model_path = 'haarcascade_frontalface_default.xml'
emotion_model_path = 'fer2013_mini_XCEPTION.102-0.66.hdf5'
emotion_labels = get_labels('fer2013')
# hyper-parameters for bounding boxes shape
frame_window = 10
emotion_offsets = (20, 40)
# loading models
face_detection = load_detection_model(detection_model_path)
emotion_classifier = load_model(emotion_model_path, compile=False)
# getting input model shapes for inference
emotion_target_size = emotion_classifier.input_shape[
1:3] # ???1,3 represent what
# starting lists for calculating modes
emotion_window = []
is_EPGA = True
image_name = '3.jpg'
main_path = 'test-images/'
text_left = -20
text_top = -20
text_gap = 20
font_size = 0.6
font_thickness = 2
image_path = main_path + image_name
stored_path = main_path + '{}-{}_'.format(is_light_net,
is_EGA) + 'result_' + image_name
if is_EGA:
emotion_labels = get_labels('ferplus')
else:
emotion_labels = get_labels('fer2013')
gender_labels = get_labels('imdb')
age_labels = get_labels('adience')
pose_labels = get_labels('aflw')
if is_single_task:
if is_light_net:
MODEL = 'mini_xception'
# emotion_path='train_weights/EmotionNetminixception/expw/freezing_true-drouout_false_1__06-0.56.hdf5'
# emotion_path = 'train_weights/EmotionNetminixception/fer2013/fer2013mini_XCEPTION.95-0.66.hdf5'
# gender_age_path=''
# gender_path='train_weights/GenderNetminixception/adience/freezing_true-drouout_false_6__12-0.86.hdf5'
# gender_path = 'train_weights/GenderNetminixception/adience/simple_CNN.81-0.96.hdf5'
# gender_path='train_weights/GenderNetminixception/imdb/freezing_true-drouout_false_1__05-0.95.hdf5'
def clicked():
label = tk.Label(root)
#load the file path
image_path = askopenfilename()
print image_path
#open the image
image_path1 = Image.open(image_path)
image_path1 = image_path1.resize((200, 200), Image.ANTIALIAS)
label.image_path1 = ImageTk.PhotoImage(image_path1)
label['image'] = label.image_path1
label.pack()
detection_model_path = '/home/akanksha/code/trained_models/detection_models/haarcascade_frontalface_default.xml'
emotion_model_path = '/home/akanksha/code/trained_models/emotion_modelsfer2013_mini_XCEPTION.86-0.43.hdf5'
emotion_labels = get_labels('fer2013')
#loading models
face_detection = cv2.CascadeClassifier(detection_model_path)
emotion_classifier = load_model(emotion_model_path, compile=False)
# getting input model shapes for inference
emotion_target_size = emotion_classifier.input_shape[1:3]
# loading images
pil_image = image.load_img(image_path, grayscale=True)
gray_image = image.img_to_array(pil_image)
gray_image = np.squeeze(gray_image)
gray_image = gray_image.astype('uint8')
faces = face_detection.detectMultiScale(gray_image, 1.3, 5)
if (len(faces)==0):
print("Please enter a valid image!")
z = Label(root, text='Please enter a valid image!',background="white")
z.pack()
var = 0
for (x, y, w, h) in faces:
gray_face = gray_image[y:y+w , x:x+h]
gray_face = cv2.resize(gray_face, (emotion_target_size))
gray_face = preprocess_input(gray_face)
gray_face = np.expand_dims(gray_face, 0)
gray_face = np.expand_dims(gray_face, -1)
emotion_proba = emotion_classifier.predict(gray_face)
print("------------------------------------------------------------------------")
print("Probabilities of each class: ")
print(" 0:angry , 1:disgust , 2:fear , 3:happy , 4:sad , 5:surprise , 6:neutral ")
print(emotion_proba)
emotion_label_arg = np.argmax(emotion_classifier.predict(gray_face))
emotion_text = emotion_labels[emotion_label_arg]
print("-------------------------------------------------------------------------")
print("Emotion class: ")
print(emotion_text)
print("--------------------------------------------------------------------------")
var = np.amax(emotion_proba)
print("Maximum probability emotion: ")
print(var)
c = Label(root, text='MAXIMUM PROBABILITY EMOTION: ',background="white")
c.pack()
d = Label(root, text=var,background="white")
d.pack()
f = Label(root, text="EMOTION LABEL: ",background="white")
f.pack()
e = Label(root, text=emotion_text,background="white")
e.pack()
var = var * 100
if(emotion_text == "anger"):
S = 2.36*(math.log(0.33*var + 1.00))
elif(emotion_text == "disguist"):
S = 7.27*(math.log(0.01*var + 1.02))
elif(emotion_text == "fear"):
S = 1.76*(math.log(1.36*var + 1.00))
elif(emotion_text == "happy"):
S = -7.56*(math.log(-0.003*var + 1.01))
elif(emotion_text == "sad"):
S = 2.85*(math.log(0.13*var + 1.01))
elif(emotion_text == "surprise"):
S = 2.45*(math.log(0.29*var + 1.00))
elif(emotion_text == "neutral"):
S = 5.05*(math.log(0.015*var + 1.016))
print("Stress Value:")
print(S)
a = Label(root, text='STRESS VALUE(range:0-9): ',background="white")
a.pack()
b = Label(root, text=S,background="white")
b.pack()
axis_array[row_arg, col_arg].axis('off')
axis_array[row_arg, col_arg].imshow(image, cmap=cmap)
axis_array[row_arg, col_arg].set_title(titles[image_arg])
image_arg = image_arg + 1
plt.tight_layout()
if __name__ == '__main__':
# from utils.data_manager import DataManager
from datasets import get_labels
from keras.models import load_model
import pickle
# dataset_name = 'fer2013'
# model_path = '../trained_models/emotion_models/simple_CNN.985-0.66.hdf5'
dataset_name = 'fer2013'
class_gender = get_labels(dataset_name)
# data_manager = DataManager(dataset_name)
# face, emotions = data_manager.get_data()
faces = pickle.load(open('face.pkl', 'rb'))
emotions = pickle.load(open('emotion.pkl', 'rb'))
pretty_imshow(plt.gca(), make_mosaic(faces[:4], 2, 2), cmap='gray')
plt.show()
"""
image_arg = 0
face = faces[image_arg:image_arg + 1]
emotion = emotions[image_arg:image_arg + 1]
display_image(face, emotion, class_decoder)
plt.show()
normal_imshow(plt.gca(), make_mosaic(faces[:4], 3, 3), cmap='gray')
from utils import extract_left_eye_center, extract_right_eye_center, get_rotation_matrix, crop_image
import importlib, cv2, json, os.path, random, glob, multiprocessing, argparse, fcntl, imghdr, dlib, re
#aligner constants
scale = 1
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor("/home/gc1569/Image_collector/face_classification/"\
"src/shape_predictor_68_face_landmarks.dat")
#gender prediction constants
detection_model_path = '/home/gc1569/Image_collector/face_classification'\
'/trained_models/detection_models/haarcascade_frontalface_default.xml'
emotion_model_path = '/home/gc1569/Image_collector/face_classification'\
'/trained_models/emotion_models/fer2013_mini_XCEPTION.102-0.66.hdf5'
gender_model_path = '/home/gc1569/Image_collector/face_classification'\
'/trained_models/gender_models/simple_CNN.81-0.96.hdf5'
emotion_labels = get_labels('fer2013')
gender_labels = get_labels('imdb')
#loading models
face_detection = load_detection_model(detection_model_path)
gender_classifier = load_model(gender_model_path, compile=False)
#getting input model shapes for inference
gender_target_size = gender_classifier.input_shape[1:3]
#hyper-parameters for bounding boxes shape
gender_offsets = (30, 60)
gender_offsets = (10, 10)
#crops images to face (in seperate iamges if more faces) and output the landmark features in a text file
def align_face(name):
input_image, output_image = source + name, destination + name
def emotion():
# parameters for loading data and images
emotion_text = ""
detection_model_path = '../trained_models/detection_models/haarcascade_frontalface_default.xml'
emotion_model_path = '../trained_models/emotion_models/fer2013_mini_XCEPTION.102-0.66.hdf5'
emotion_labels = get_labels('fer2013')
# hyper-parameters for bounding boxes shape
frame_window = 10
emotion_offsets = (20, 40)
# loading models
face_detection = load_detection_model(detection_model_path)
emotion_classifier = load_model(emotion_model_path, compile=False)
# getting input model shapes for inference
emotion_target_size = emotion_classifier.input_shape[1:3]
# starting lists for calculating modes
emotion_window = []
# starting video streaming
cv2.namedWindow('window_frame')
video_capture = cv2.VideoCapture(0)
while True:
bgr_image = video_capture.read()[1]
gray_image = cv2.cvtColor(bgr_image, cv2.COLOR_BGR2GRAY)
rgb_image = cv2.cvtColor(bgr_image, cv2.COLOR_BGR2RGB)
faces = detect_faces(face_detection, gray_image)
for face_coordinates in faces:
x1, x2, y1, y2 = apply_offsets(face_coordinates, emotion_offsets)
gray_face = gray_image[y1:y2, x1:x2]
try:
gray_face = cv2.resize(gray_face, (emotion_target_size))
except:
continue
gray_face = preprocess_input(gray_face, True)
gray_face = np.expand_dims(gray_face, 0)
gray_face = np.expand_dims(gray_face, -1)
emotion_prediction = emotion_classifier.predict(gray_face)
emotion_probability = np.max(emotion_prediction)
emotion_label_arg = np.argmax(emotion_prediction)
emotion_text = emotion_labels[emotion_label_arg]
emotion_window.append(emotion_text)
if len(emotion_window) > frame_window:
emotion_window.pop(0)
try:
emotion_mode = mode(emotion_window)
except:
continue
if emotion_text == 'angry':
color = emotion_probability * np.asarray((255, 0, 0))
elif emotion_text == 'sad':
color = emotion_probability * np.asarray((0, 0, 255))
elif emotion_text == 'happy':
color = emotion_probability * np.asarray((255, 255, 0))
elif emotion_text == 'surprise':
color = emotion_probability * np.asarray((0, 255, 255))
else:
color = emotion_probability * np.asarray((0, 255, 0))
color = color.astype(int)
color = color.tolist()
draw_bounding_box(face_coordinates, rgb_image, color)
draw_text(face_coordinates, rgb_image, emotion_mode, color, 0, -45,
1, 1)
if (emotion_text):
return emotion_text
break
bgr_image = cv2.cvtColor(rgb_image, cv2.COLOR_RGB2BGR)
cv2.imshow('window_frame', bgr_image)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
if pvalue < 0.05:
# rejeitando a hipotese de que nao existe diferenca entre
# os classificadores
rate_matrix = rate_matrix.melt(var_name='groups', value_name='values')
nemenyi_results = posthoc_nemenyi(rate_matrix, val_col='values', group_col='groups')
print('teste de nemenyi')
for i in range(0, nemenyi_results.shape[0]):
print(nemenyi_results.iloc[i])
data_set_classes = ['BRICKFACE', 'SKY', 'FOLIAGE', 'CEMENT', 'WINDOW', 'PATH', 'GRASS']
training_set = datasets.read_set('segmentation.test')
training_set = datasets.normalize_data_set(training_set)
training_shape_view, training_rgb_view = datasets.split_views(training_set)
data_set_labels = datasets.get_labels(training_set)
# removendo as colunas 3, 4 e 5 do training set original e da shape view
# pois todas tem variabilidade muito baixa (var -> 0)
training_set = training_set.drop(columns=['REGION-PIXEL-COUNT',
'SHORT-LINE-DENSITY-5',
'SHORT-LINE-DENSITY-2'])
training_shape_view = training_shape_view.drop(columns=['REGION-PIXEL-COUNT',
'SHORT-LINE-DENSITY-5',
'SHORT-LINE-DENSITY-2'])
# melhores ks obtidos para o conjunto de treinamento e a vies
# isto foi obtido por validacao cruzada (validate_k do arquivo knn.py)
training_set_k = 15
shape_view_k = 19
rgb_view_k = 15
axis_array[row_arg, col_arg].imshow(image, cmap=cmap)
axis_array[row_arg, col_arg].set_title(titles[image_arg])
image_arg = image_arg + 1
plt.tight_layout()
if __name__ == '__main__':
#from utils.data_manager import DataManager
from datasets import get_labels
from keras.models import load_model
import pickle
#dataset_name = 'fer2013'
#model_path = '../trained_models/emotion_models/simple_CNN.985-0.66.hdf5'
dataset_name = 'fer2018'
class_decoder = get_labels(dataset_name)
#data_manager = DataManager(dataset_name)
#faces, emotions = data_manager.get_data()
faces = pickle.load(open('faces.pkl', 'rb'))
emotions = pickle.load(open('emotions.pkl', 'rb'))
pretty_imshow(plt.gca(), make_mosaic(faces[:4], 2, 2), cmap='gray')
plt.show()
"""
image_arg = 0
face = faces[image_arg:image_arg + 1]
emotion = emotions[image_arg:image_arg + 1]
display_image(face, emotion, class_decoder)
plt.show()
normal_imshow(plt.gca(), make_mosaic(faces[:4], 3, 3), cmap='gray')
plt.show()