aistudio上传文件夹 把整个文件压缩 在主文件里面 添加代码行import os
!unzip genki4k.zip（压缩包文件名）
main.py
注意点与心得：1.一开始用百度飞桨跑 结果忘记把文件上传导致路径一直报错 后来发现aistudio已经不支持使用tensorflow和pytorch框架 无奈又去使用pycharm
2.安装dlib opencv tensorflow
使用pip命令进入anaconda prompt频繁报错
参考：https://pywin.blog.csdn.net/article/details/114663880
3.安装dlib
https://blog.csdn.net/qq_40628213/article/details/121874659?ops_request_misc=&request_id=&biz_id=102&utm_term=python3.6%E5%AE%89%E8%A3%85dlib&utm_medium=distribute.pc_search_result.none-task-blog-2blogsobaiduweb~default-0-121874659.nonecase&spm=1018.2226.3001.4450
4.安装的时候要切换到d盘 cd到annaconda/script路径下
5.期间有一个生成文件 需要进行异常抛出 否则会报错
import dlib # 人脸识别的库dlib
import numpy as np # 数据处理的库numpy
import cv2 # 图像处理的库OpenCv
import os
# dlib预测器
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor('./genki4k/shape_predictor_68_face_landmarks.dat')
# 读取图像的路径
path_read = "./genki4k/files"
num = 0
for file_name in os.listdir(path_read):
# aa是图片的全路径
aa = (path_read + "/" + file_name)
# 读入的图片的路径中含非英文
img = cv2.imdecode(np.fromfile(aa, dtype=np.uint8), cv2.IMREAD_UNCHANGED)
# 获取图片的宽高
img_shape = img.shape
img_height = img_shape[0]
img_width = img_shape[1]
# 用来存储生成的单张人脸的路径
path_save = "./genki4k/files1"
# dlib检测
dets = detector(img, 1)
# print("人脸数：", len(dets))
for k, d in enumerate(dets):
if len(dets)>1:
continue
num=num+1
# 计算矩形大小
# (x,y), (宽度width, 高度height)
pos_start = tuple([d.left(), d.top()])
pos_end = tuple([d.right(), d.bottom()])
# 计算矩形框大小
height = d.bottom()-d.top()
width = d.right()-d.left()
# 根据人脸大小生成空的图像
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)
cv2.imencode('.jpg', img_blank)[1].tofile(path_save+"\\"+"file"+str(num)+".jpg") # 正确方法
train.py
import os, shutil
#创建模型
from keras import layers
from keras import models
from keras import optimizers
from tensorflow import optimizers
import matplotlib.pyplot as plt
# This is module with image preprocessing utilities
from keras.preprocessing import image
from keras.preprocessing.image import ImageDataGenerator
# 原始数据集路径
original_dataset_dir = './genki4k/files'
# 新的数据集
base_dir = './genki4k/files1'
# os.mkdir(base_dir)
try:
os.mkdir(base_dir)
except OSError:
pass
# # 训练图像、验证图像、测试图像的目录
train_dir = os.path.join(base_dir, 'train')
try:
os.mkdir(train_dir)
except OSError:
pass
validation_dir = os.path.join(base_dir, 'validation')
# os.mkdir(validation_dir)
try:
os.mkdir(validation_dir)
except OSError:
pass
test_dir = os.path.join(base_dir, 'test')
# os.mkdir(test_dir)
try:
os.mkdir(test_dir)
except OSError:
pass
# 笑训练图片所在目录
train_smile_dir = os.path.join(train_dir, 'smile')
try:
os.mkdir(train_smile_dir)
except OSError:
pass
# 不笑训练图片所在目录
train_unsmile_dir = os.path.join(train_dir, 'unsmile')
try:
os.mkdir(train_unsmile_dir)
except OSError:
pass
# 笑验证图片所在目录
validation_smile_dir = os.path.join(validation_dir, 'smile')
try:
os.mkdir(validation_smile_dir)
except OSError:
pass
# 不笑验证图片所在目录
validation_unsmile_dir = os.path.join(validation_dir, 'unsmile')
try:
os.mkdir(validation_unsmile_dir)
except OSError:
pass
# 笑测试图片所在目录
test_smile_dir = os.path.join(test_dir, 'smile')
try:
os.mkdir(test_smile_dir)
except OSError:
pass
# 不笑测试图片所在目录
test_unsmile_dir = os.path.join(test_dir, 'unsmile')
# os.mkdir(test_dogs_dir)
try:
os.mkdir(test_smile_dir)
except OSError:
pass
# 复制1000张笑脸图片到train_c_dir
fnames = ['file00{}.jpg'.format(i) for i in range(1, 99)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(train_smile_dir, fname)
shutil.copyfile(src, dst)
fnames = ['file0{}.jpg'.format(i) for i in range(100, 900)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(train_smile_dir, fname)
shutil.copyfile(src, dst)
fnames = ['file0{}.jpg'.format(i) for i in range(900, 999)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(validation_smile_dir, fname)
shutil.copyfile(src, dst)
fnames = ['file{}.jpg'.format(i) for i in range(1000, 1350)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(validation_smile_dir, fname)
shutil.copyfile(src, dst)
# Copy next 500 cat images to test_smile_dir
fnames = ['file{}.jpg'.format(i) for i in range(1350, 1800)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(test_smile_dir, fname)
shutil.copyfile(src, dst)
fnames = ['file{}.jpg'.format(i) for i in range(2127, 3000)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(train_smile_dir, fname)
shutil.copyfile(src, dst)
# Copy next 500 unsmile images to validation_unsmile_dir
fnames = ['file{}.jpg'.format(i) for i in range(3000, 3878)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(validation_unsmile_dir, fname)
shutil.copyfile(src, dst)
# Copy next 500 unsmile images to test_unsmile_dir
fnames = ['file{}.jpg'.format(i) for i in range(3000, 3878)]
for fname in fnames:
src = os.path.join(original_dataset_dir, fname)
dst = os.path.join(test_unsmile_dir, fname)
shutil.copyfile(src, dst)
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.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
model.summary()#查看
#归一化
model.compile(loss='binary_crossentropy',
optimizer=optimizers.RMSprop(lr=1e-4),
metrics=['acc'])
train_datagen = ImageDataGenerator(rescale=1./255)
validation_datagen=ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow_from_directory(
# 目标文件目录
train_dir,
#所有图片的size必须是150x150
target_size=(150, 150),
batch_size=20,
# 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=20,
class_mode='binary')
test_generator = test_datagen.flow_from_directory(test_dir,
target_size=(150, 150),
batch_size=20,
class_mode='binary')
for data_batch, labels_batch in train_generator:
print('data batch shape:', data_batch.shape)
print('labels batch shape:', labels_batch)
break
# 训练模型
history = model.fit_generator(
train_generator,
steps_per_epoch=100,
epochs=30,
validation_data=validation_generator,
validation_steps=50)
# 保存模型
model.save('smile_and_unsmile_small_1.h5')
# 在训练和验证数据上绘制模型的损失和准确性
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']
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()
# 数据增强
datagen = ImageDataGenerator(
rotation_range=40,
width_shift_range=0.2,
height_shift_range=0.2,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True,
fill_mode='nearest')
# This is module with image preprocessing utilities
from keras.preprocessing import image
fnames = [os.path.join(train_smile_dir, fname) for fname in os.listdir(train_smile_dir)]
# We pick one image to "augment"
img_path = fnames[3]
# Read the image and resize it
img = image.load_img(img_path, target_size=(150, 150))
# Convert it to a Numpy array with shape (150, 150, 3)
x = image.img_to_array(img)
# Reshape it to (1, 150, 150, 3)
x = x.reshape((1,) + x.shape)
# The .flow() command below generates batches of randomly transformed images.
# It will loop indefinitely, so we need to `break` the loop at some point!
i = 0
for batch in datagen.flow(x, batch_size=1):
plt.figure(i)
imgplot = plt.imshow(image.array_to_img(batch[0]))
i += 1
if i % 4 == 0:
break
plt.show()
# 添加drop层
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'])
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,)
# 训练
# Note that the validation data should not be augmented!
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=100,
epochs=100,
validation_data=validation_generator,
validation_steps=50)
`
扫一扫
1908
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
