溫馨提示×
您好，登錄后才能下訂單哦！
點擊 登錄注冊 即表示同意《億速云用戶服務條款》
您好，登錄后才能下訂單哦！
這篇文章主要介紹Python基于keras訓練如何實現微笑識別，文中介紹的非常詳細，具有一定的參考價值，感興趣的小伙伴們一定要看完！
實驗數據來自genki4k
提取含有完整人臉的圖片
def init_file():
num = 0
bar = tqdm(os.listdir(read_path))
for file_name in bar:
bar.desc = "預處理圖片: "
# a圖片的全路徑
img_path = (read_path + "/" + file_name)
# 讀入的圖片的路徑中含非英文
img = cv2.imdecode(np.fromfile(img_path, dtype=np.uint8), cv2.IMREAD_UNCHANGED)
# 獲取圖片的寬高
img_shape = img.shape
img_height = img_shape[0]
img_width = img_shape[1]
# 用來存儲生成的單張人臉的路徑
# dlib檢測
dets = detector(img, 1)
for k, d in enumerate(dets):
if len(dets) > 1:
continue
num += 1
# 計算矩形大小
# (x,y), (寬度width, 高度height)
# pos_start = tuple([d.left(), d.top()])
# pos_end = tuple([d.right(), d.bottom()])
# 計算矩形框大小
height = d.bottom() - d.top()
width = d.right() - d.left()
# 根據人臉大小生成空的圖像
img_blank = np.zeros((height, width, 3), np.uint8)
for i in range(height):
if d.top() + i >= img_height: # 防止越界
continue
for j in range(width):
if d.left() + j >= img_width: # 防止越界
continue
img_blank[i][j] = img[d.top() + i][d.left() + j]
img_blank = cv2.resize(img_blank, (200, 200), interpolation=cv2.INTER_CUBIC)
# 保存圖片
cv2.imencode('.jpg', img_blank)[1].tofile(save_path + "/" + "file" + str(num) + ".jpg")
logging.info("一共", len(os.listdir(read_path)), "個樣本")
logging.info("有效樣本", num)# 創建網絡 def create_model(): model = models.Sequential() model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(150, 150, 3))) model.add(layers.MaxPooling2D((2, 2))) model.add(layers.Conv2D(64, (3, 3), activation='relu')) model.add(layers.MaxPooling2D((2, 2))) model.add(layers.Conv2D(128, (3, 3), activation='relu')) model.add(layers.MaxPooling2D((2, 2))) model.add(layers.Conv2D(128, (3, 3), activation='relu')) model.add(layers.MaxPooling2D((2, 2))) model.add(layers.Flatten()) model.add(layers.Dropout(0.5)) model.add(layers.Dense(512, activation='relu')) model.add(layers.Dense(1, activation='sigmoid')) model.compile(loss='binary_crossentropy', optimizer=optimizers.RMSprop(lr=1e-4), metrics=['acc']) return model
# 訓練模型 def train_model(model): # 歸一化處理 train_datagen = ImageDataGenerator( rescale=1. / 255, rotation_range=40, width_shift_range=0.2, height_shift_range=0.2, shear_range=0.2, zoom_range=0.2, horizontal_flip=True, ) test_datagen = ImageDataGenerator(rescale=1. / 255) train_generator = train_datagen.flow_from_directory( # This is the target directory train_dir, # All images will be resized to 150x150 target_size=(150, 150), batch_size=32, # Since we use binary_crossentropy loss, we need binary labels class_mode='binary') validation_generator = test_datagen.flow_from_directory( validation_dir, target_size=(150, 150), batch_size=32, class_mode='binary') history = model.fit_generator( train_generator, steps_per_epoch=60, epochs=12, validation_data=validation_generator, validation_steps=30) # 保存模型 save_path = "../output/model" if not os.path.exists(save_path): os.makedirs(save_path) model.save(save_path + "/smileDetect.h6") return history
準確率
丟失率
訓練過程
通過讀取攝像頭內容進行預測
def rec(img):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
dets = detector(gray, 1)
if dets is not None:
for face in dets:
left = face.left()
top = face.top()
right = face.right()
bottom = face.bottom()
cv2.rectangle(img, (left, top), (right, bottom), (0, 255, 0), 2)
img1 = cv2.resize(img[top:bottom, left:right], dsize=(150, 150))
img1 = cv2.cvtColor(img1, cv2.COLOR_BGR2RGB)
img1 = np.array(img1) / 255.
img_tensor = img1.reshape(-1, 150, 150, 3)
prediction = model.predict(img_tensor)
if prediction[0][0] > 0.5:
result = 'unsmile'
else:
result = 'smile'
cv2.putText(img, result, (left, top), font, 2, (0, 255, 0), 2, cv2.LINE_AA)
cv2.imshow('Video', img)
while video.isOpened():
res, img_rd = video.read()
if not res:
break
rec(img_rd)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
import dlib # 人臉識別的庫dlib
import numpy as np # 數據處理的庫numpy
import cv2 # 圖像處理的庫OpenCv
import os
import shutil
from tqdm import tqdm
import logging
# dlib預測器
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor('../resources/shape_predictor_68_face_landmarks.dat')
# 原圖片路徑
read_path = "../resources/genki4k/files"
# 提取人臉存儲路徑
save_path = "../output/genki4k/files"
if not os.path.exists(save_path):
os.makedirs(save_path)
# 新的數據集
data_dir = '../resources/data'
if not os.path.exists(data_dir):
os.makedirs(data_dir)
# 訓練集
train_dir = data_dir + "/train"
if not os.path.exists(train_dir):
os.makedirs(train_dir)
# 驗證集
validation_dir = os.path.join(data_dir, 'validation')
if not os.path.exists(validation_dir):
os.makedirs(validation_dir)
# 測試集
test_dir = os.path.join(data_dir, 'test')
if not os.path.exists(test_dir):
os.makedirs(test_dir)
# 初始化訓練數據
def init_data(file_list):
# 如果不存在文件夾則新建
for file_path in file_list:
if not os.path.exists(file_path):
os.makedirs(file_path)
# 存在則清空里面所有數據
else:
for i in os.listdir(file_path):
path = os.path.join(file_path, i)
if os.path.isfile(path):
os.remove(path)
def init_file():
num = 0
bar = tqdm(os.listdir(read_path))
for file_name in bar:
bar.desc = "預處理圖片: "
# a圖片的全路徑
img_path = (read_path + "/" + file_name)
# 讀入的圖片的路徑中含非英文
img = cv2.imdecode(np.fromfile(img_path, dtype=np.uint8), cv2.IMREAD_UNCHANGED)
# 獲取圖片的寬高
img_shape = img.shape
img_height = img_shape[0]
img_width = img_shape[1]
# 用來存儲生成的單張人臉的路徑
# dlib檢測
dets = detector(img, 1)
for k, d in enumerate(dets):
if len(dets) > 1:
continue
num += 1
# 計算矩形大小
# (x,y), (寬度width, 高度height)
# pos_start = tuple([d.left(), d.top()])
# pos_end = tuple([d.right(), d.bottom()])
# 計算矩形框大小
height = d.bottom() - d.top()
width = d.right() - d.left()
# 根據人臉大小生成空的圖像
img_blank = np.zeros((height, width, 3), np.uint8)
for i in range(height):
if d.top() + i >= img_height: # 防止越界
continue
for j in range(width):
if d.left() + j >= img_width: # 防止越界
continue
img_blank[i][j] = img[d.top() + i][d.left() + j]
img_blank = cv2.resize(img_blank, (200, 200), interpolation=cv2.INTER_CUBIC)
# 保存圖片
cv2.imencode('.jpg', img_blank)[1].tofile(save_path + "/" + "file" + str(num) + ".jpg")
logging.info("一共", len(os.listdir(read_path)), "個樣本")
logging.info("有效樣本", num)
# 劃分數據集
def divide_data(file_path, message, begin, end):
files = ['file{}.jpg'.format(i) for i in range(begin, end)]
bar = tqdm(files)
bar.desc = message
for file in bar:
src = os.path.join(save_path, file)
dst = os.path.join(file_path, file)
shutil.copyfile(src, dst)
if __name__ == "__main__":
init_file()
positive_train_dir = os.path.join(train_dir, 'smile')
negative_train_dir = os.path.join(train_dir, 'unSmile')
positive_validation_dir = os.path.join(validation_dir, 'smile')
negative_validation_dir = os.path.join(validation_dir, 'unSmile')
positive_test_dir = os.path.join(test_dir, 'smile')
negative_test_dir = os.path.join(test_dir, 'unSmile')
file_list = [positive_train_dir, positive_validation_dir, positive_test_dir,
negative_train_dir, negative_validation_dir, negative_test_dir]
init_data(file_list)
divide_data(positive_train_dir, "劃分訓練集正樣本", 1, 1001)
divide_data(negative_train_dir, "劃分訓練集負樣本", 2200, 3200)
divide_data(positive_validation_dir, "劃分驗證集正樣本", 1000, 1500)
divide_data(negative_validation_dir, "劃分驗證集負樣本", 3000, 3500)
divide_data(positive_test_dir, "劃分測試集正樣本", 1500, 2000)
divide_data(negative_test_dir, "劃分測試集負樣本", 2800, 3500)import os
from keras import layers
from keras import models
from tensorflow import optimizers
import matplotlib.pyplot as plt
from keras.preprocessing.image import ImageDataGenerator
train_dir = "../resources/data/train"
validation_dir = "../resources/data/validation"
# 創建網絡
def create_model():
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dropout(0.5))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy',
optimizer=optimizers.RMSprop(lr=1e-4),
metrics=['acc'])
return model
# 訓練模型
def train_model(model):
# 歸一化處理
train_datagen = ImageDataGenerator(
rescale=1. / 255,
rotation_range=40,
width_shift_range=0.2,
height_shift_range=0.2,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True, )
test_datagen = ImageDataGenerator(rescale=1. / 255)
train_generator = train_datagen.flow_from_directory(
# This is the target directory
train_dir,
# All images will be resized to 150x150
target_size=(150, 150),
batch_size=32,
# Since we use binary_crossentropy loss, we need binary labels
class_mode='binary')
validation_generator = test_datagen.flow_from_directory(
validation_dir,
target_size=(150, 150),
batch_size=32,
class_mode='binary')
history = model.fit_generator(
train_generator,
steps_per_epoch=60,
epochs=12,
validation_data=validation_generator,
validation_steps=30)
# 保存模型
save_path = "../output/model"
if not os.path.exists(save_path):
os.makedirs(save_path)
model.save(save_path + "/smileDetect.h6")
return history
# 展示訓練結果
def show_results(history):
# 數據增強過后的訓練集與驗證集的精確度與損失度的圖形
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']
# 繪制結果
epochs = range(len(acc))
plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()
plt.figure()
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
plt.show()
if __name__ == "__main__":
model = create_model()
history = train_model(model)
show_results(history)import os
from keras import layers
from keras import models
from tensorflow import optimizers
import matplotlib.pyplot as plt
from keras.preprocessing.image import ImageDataGenerator
train_dir = "../resources/data/train"
validation_dir = "../resources/data/validation"
# 創建網絡
# 檢測視頻或者攝像頭中的人臉
import cv2
from keras.preprocessing import image
from keras.models import load_model
import numpy as np
import dlib
from PIL import Image
model = load_model('../output/model/smileDetect.h6')
detector = dlib.get_frontal_face_detector()
video = cv2.VideoCapture(0)
font = cv2.FONT_HERSHEY_SIMPLEX
def rec(img):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
dets = detector(gray, 1)
if dets is not None:
for face in dets:
left = face.left()
top = face.top()
right = face.right()
bottom = face.bottom()
cv2.rectangle(img, (left, top), (right, bottom), (0, 255, 0), 2)
img1 = cv2.resize(img[top:bottom, left:right], dsize=(150, 150))
img1 = cv2.cvtColor(img1, cv2.COLOR_BGR2RGB)
img1 = np.array(img1) / 255.
img_tensor = img1.reshape(-1, 150, 150, 3)
prediction = model.predict(img_tensor)
if prediction[0][0] > 0.5:
result = 'unsmile'
else:
result = 'smile'
cv2.putText(img, result, (left, top), font, 2, (0, 255, 0), 2, cv2.LINE_AA)
cv2.imshow('Video', img)
while video.isOpened():
res, img_rd = video.read()
if not res:
break
rec(img_rd)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
video.release()
cv2.destroyAllWindows()以上是“Python基于keras訓練如何實現微笑識別”這篇文章的所有內容，感謝各位的閱讀！希望分享的內容對大家有幫助，更多相關知識，歡迎關注億速云行業資訊頻道！
免責聲明：本站發布的內容（圖片、視頻和文字）以原創、轉載和分享為主，文章觀點不代表本網站立場，如果涉及侵權請聯系站長郵箱：is@yisu.com進行舉報，并提供相關證據，一經查實，將立刻刪除涉嫌侵權內容。
