Phân loại hình ảnh với Tensorflow - Bài 2 chạy code Python trên Windows 10

Đây là bài thứ 2 trong chuỗi bài viết về Image Classification với thư viện Tensorflow. Sau khi đọc và làm theo Bài 1 thử nghiệm độ chính xác thì các bạn đã giải thuật này rất là hiệu quả.

Bài này sẽ hướng dẫn cách thực thi code Python trên Windows 10 do copy code trên Notebook về không sử dụng được ngay. Khi viết bài này mình sử dụng GPU NVIDIA RTX 2060 để tăng tốc training và predict. Nếu máy bạn không có GPU hoặc laptop thì chạy rất là chậm, có thể gấp 6-20 lần thời gian.

Chuẩn bị

Python 3.7.3: miễn là Python 3.6 trở lên là được, cài xong thì restart máy
CUDA 10.1 trở lên

Cài đặt Tensorflow 2.3 vì máy mình cài CUDA 10.1, nếu bạn muốn cài Tensorflow 2.6 trở lên thì phải cài CUDA 11.1.
pip install tensorflow-gpu==2.3

Nếu máy không có GPU thì cài bằng lệnh
pip install tensorflow==2.3

Cài các package khác:
matplotlib==3.4.3 numpy==1.18.5 Pillow==8.3.2

Download hình ảnh về và giải nén vào folder flower_photos, trong folder này có 5 folder chứa 5 loài hoa
https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz

Source code

Save nội dung bên dưới vào file image_classification.py

[bg_collapse view=”button-orange” color=”#4a4949″ expand_text=”image_classification.py” collapse_text=”image_classification.py” ]

import sys
import matplotlib.pyplot as plt
import numpy as np
import os
import PIL
import tensorflow as tf
import fnmatch

tf.config.experimental.list_physical_devices('GPU')

from tensorflow import keras
from tensorflow.keras import layers
from tensorflow.keras.models import Sequential

def GetFileName(filePath):
    fileName = os.path.basename(filePath)
    return fileName
    
def GetDirName(fullpath):    
    return GetFileName(fullpath)

def FindFileInDir(dirPath, pattern, fullPath=True, searchChildDir=False):
    result = []
    for fullpathDir, subdirs, files in os.walk(dirPath):
        isChildDir = len(fullpathDir) > len(dirPath)
        if(not searchChildDir and isChildDir):
            continue
        currentDirName = GetDirName(fullpathDir)

        for fileName in files:            
            if fnmatch.fnmatch(fileName, pattern):
                if(fullPath):
                    result.append(os.path.join(fullpathDir, fileName))
                else:
                    result.append(os.path.join(currentDirName, fileName))

    return result


data_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "flower_photos")

imgPaths = FindFileInDir(data_dir, "*.jpg", False, True)
image_count = len(imgPaths)

# roses = FindFileInDir(os.path.join(data_dir, 'roses'), "*.jpg")
# print(str(roses[0]))
# img = PIL.Image.open(str(roses[0]))
#img.show()

batch_size = 32
img_height = 180
img_width = 180

train_ds = tf.keras.preprocessing.image_dataset_from_directory(
    data_dir,
    validation_split=0.2,
    subset="training",
    seed=123,
    image_size=(img_height, img_width),
    batch_size=batch_size)

val_ds = tf.keras.preprocessing.image_dataset_from_directory(
    data_dir,
    validation_split=0.2,
    subset="validation",
    seed=123,
    image_size=(img_height, img_width),
    batch_size=batch_size)

class_names = train_ds.class_names
print(class_names)


plt.figure(figsize=(10, 10))
for images, labels in train_ds.take(1):
    for i in range(9):
        ax = plt.subplot(3, 3, i + 1)
        plt.imshow(images[i].numpy().astype("uint8"))
        plt.title(class_names[labels[i]])
        plt.axis("off")


for image_batch, labels_batch in train_ds:
    print(image_batch.shape)
    print(labels_batch.shape)
    break

AUTOTUNE = tf.data.experimental.AUTOTUNE

train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)
val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)

normalization_layer = layers.experimental.preprocessing.Rescaling(1./255)


normalized_ds = train_ds.map(lambda x, y: (normalization_layer(x), y))
image_batch, labels_batch = next(iter(normalized_ds))
first_image = image_batch[0]
# Notice the pixels values are now in `[0,1]`.
print(np.min(first_image), np.max(first_image)) 


num_classes = 5

model = Sequential([
    layers.experimental.preprocessing.Rescaling(1./255, input_shape=(img_height, img_width, 3)),
    layers.Conv2D(16, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(32, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(64, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Flatten(),
    layers.Dense(128, activation='relu'),
    layers.Dense(num_classes)
])



model.compile(optimizer='adam',
                loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
                metrics=['accuracy'])
            
model.summary()

epochs=10
history = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=epochs
)

acc = history.history['accuracy']
val_acc = history.history['val_accuracy']

loss = history.history['loss']
val_loss = history.history['val_loss']

epochs_range = range(epochs)

plt.figure(figsize=(8, 8))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, acc, label='Training Accuracy')
plt.plot(epochs_range, val_acc, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, label='Training Loss')
plt.plot(epochs_range, val_loss, label='Validation Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
#plt.show()


data_augmentation = keras.Sequential(
    [
        layers.experimental.preprocessing.RandomFlip("horizontal", 
                                                    input_shape=(img_height, 
                                                                img_width,
                                                                3)),
        layers.experimental.preprocessing.RandomRotation(0.1),
        layers.experimental.preprocessing.RandomZoom(0.1),
    ]
)

plt.figure(figsize=(10, 10))
for images, _ in train_ds.take(1):
    for i in range(9):
        augmented_images = data_augmentation(images)
        ax = plt.subplot(3, 3, i + 1)
        plt.imshow(augmented_images[0].numpy().astype("uint8"))
        plt.axis("off")

model = Sequential([
    data_augmentation,
    layers.experimental.preprocessing.Rescaling(1./255),
    layers.Conv2D(16, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(32, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(64, 3, padding='same', activation='relu'),
    layers.MaxPooling2D(),
    layers.Dropout(0.2),
    layers.Flatten(),
    layers.Dense(128, activation='relu'),
    layers.Dense(num_classes)
])


model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

model.summary()

epochs = 15
history = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=epochs
)

acc = history.history['accuracy']
val_acc = history.history['val_accuracy']

loss = history.history['loss']
val_loss = history.history['val_loss']

epochs_range = range(epochs)

plt.figure(figsize=(8, 8))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, acc, label='Training Accuracy')
plt.plot(epochs_range, val_acc, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, label='Training Loss')
plt.plot(epochs_range, val_loss, label='Validation Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
#plt.show()

sunflower_url = "https://storage.googleapis.com/download.tensorflow.org/example_images/592px-Red_sunflower.jpg"
sunflower_path = tf.keras.utils.get_file('Red_sunflower', origin=sunflower_url)

img = keras.preprocessing.image.load_img(
    sunflower_path, target_size=(img_height, img_width)
)
img_array = keras.preprocessing.image.img_to_array(img)
img_array = tf.expand_dims(img_array, 0) # Create a batch

model.save("model_flower.h5")

new_model = tf.keras.models.load_model('model_flower.h5')

predictions = new_model.predict(img_array)
score = tf.nn.softmax(predictions[0])

print(
    "This image most likely belongs to {} with a {:.2f} percent confidence."
    .format(class_names[np.argmax(score)], 100 * np.max(score))
)
os.remove(sunflower_path)

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

import sys

import matplotlib.pyplot as plt

import numpy as np

import os

import PIL

import tensorflow as tf

import fnmatch

tf.config.experimental.list_physical_devices('GPU')

from tensorflow import keras

from tensorflow.keras import layers

from tensorflow.keras.models import Sequential

def GetFileName(filePath):

fileName = os.path.basename(filePath)

return fileName

def GetDirName(fullpath):

return GetFileName(fullpath)

def FindFileInDir(dirPath, pattern, fullPath=True, searchChildDir=False):

result = []

for fullpathDir, subdirs, files in os.walk(dirPath):

isChildDir = len(fullpathDir) > len(dirPath)

if(not searchChildDir and isChildDir):

continue

currentDirName = GetDirName(fullpathDir)

for fileName in files:

if fnmatch.fnmatch(fileName, pattern):

if(fullPath):

result.append(os.path.join(fullpathDir, fileName))

else:

result.append(os.path.join(currentDirName, fileName))

return result

data_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "flower_photos")

imgPaths = FindFileInDir(data_dir, "*.jpg", False, True)

image_count = len(imgPaths)

# roses = FindFileInDir(os.path.join(data_dir, 'roses'), "*.jpg")

# print(str(roses[0]))

# img = PIL.Image.open(str(roses[0]))

#img.show()

batch_size = 32

img_height = 180

img_width = 180

train_ds = tf.keras.preprocessing.image_dataset_from_directory(

data_dir,

validation_split=0.2,

subset="training",

seed=123,

image_size=(img_height, img_width),

batch_size=batch_size)

val_ds = tf.keras.preprocessing.image_dataset_from_directory(

data_dir,

validation_split=0.2,

subset="validation",

seed=123,

image_size=(img_height, img_width),

batch_size=batch_size)

class_names = train_ds.class_names

print(class_names)

plt.figure(figsize=(10, 10))

for images, labels in train_ds.take(1):

for i in range(9):

ax = plt.subplot(3, 3, i + 1)

plt.imshow(images[i].numpy().astype("uint8"))

plt.title(class_names[labels[i]])

plt.axis("off")

for image_batch, labels_batch in train_ds:

print(image_batch.shape)

print(labels_batch.shape)

break

AUTOTUNE = tf.data.experimental.AUTOTUNE

train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)

val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)

normalization_layer = layers.experimental.preprocessing.Rescaling(1./255)

normalized_ds = train_ds.map(lambda x, y: (normalization_layer(x), y))

image_batch, labels_batch = next(iter(normalized_ds))

first_image = image_batch[0]

# Notice the pixels values are now in `[0,1]`.

print(np.min(first_image), np.max(first_image))

num_classes = 5

model = Sequential([

layers.experimental.preprocessing.Rescaling(1./255, input_shape=(img_height, img_width, 3)),

layers.Conv2D(16, 3, padding='same', activation='relu'),

layers.MaxPooling2D(),

layers.Conv2D(32, 3, padding='same', activation='relu'),

layers.MaxPooling2D(),

layers.Conv2D(64, 3, padding='same', activation='relu'),

layers.MaxPooling2D(),

layers.Flatten(),

layers.Dense(128, activation='relu'),

layers.Dense(num_classes)

])

model.compile(optimizer='adam',

loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),

metrics=['accuracy'])

model.summary()

epochs=10

history = model.fit(

train_ds,

validation_data=val_ds,

epochs=epochs

)

acc = history.history['accuracy']

val_acc = history.history['val_accuracy']

loss = history.history['loss']

val_loss = history.history['val_loss']

epochs_range = range(epochs)

plt.figure(figsize=(8, 8))

plt.subplot(1, 2, 1)

plt.plot(epochs_range, acc, label='Training Accuracy')

plt.plot(epochs_range, val_acc, label='Validation Accuracy')

plt.legend(loc='lower right')

plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)

plt.plot(epochs_range, loss, label='Training Loss')

plt.plot(epochs_range, val_loss, label='Validation Loss')

plt.legend(loc='upper right')

plt.title('Training and Validation Loss')

#plt.show()

data_augmentation = keras.Sequential(

[

layers.experimental.preprocessing.RandomFlip("horizontal",

input_shape=(img_height,

img_width,

3)),

layers.experimental.preprocessing.RandomRotation(0.1),

layers.experimental.preprocessing.RandomZoom(0.1),

]

)

plt.figure(figsize=(10, 10))

for images, _ in train_ds.take(1):

for i in range(9):

augmented_images = data_augmentation(images)

ax = plt.subplot(3, 3, i + 1)

plt.imshow(augmented_images[0].numpy().astype("uint8"))

plt.axis("off")

model = Sequential([

data_augmentation,

layers.experimental.preprocessing.Rescaling(1./255),

layers.Conv2D(16, 3, padding='same', activation='relu'),

layers.MaxPooling2D(),

layers.Conv2D(32, 3, padding='same', activation='relu'),

layers.MaxPooling2D(),

layers.Conv2D(64, 3, padding='same', activation='relu'),

layers.MaxPooling2D(),

layers.Dropout(0.2),

layers.Flatten(),

layers.Dense(128, activation='relu'),

layers.Dense(num_classes)

])

model.compile(optimizer='adam',

loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),

metrics=['accuracy'])

model.summary()

epochs = 15

history = model.fit(

train_ds,

validation_data=val_ds,

epochs=epochs

)

acc = history.history['accuracy']

val_acc = history.history['val_accuracy']

loss = history.history['loss']

val_loss = history.history['val_loss']

epochs_range = range(epochs)

plt.figure(figsize=(8, 8))

plt.subplot(1, 2, 1)

plt.plot(epochs_range, acc, label='Training Accuracy')

plt.plot(epochs_range, val_acc, label='Validation Accuracy')

plt.legend(loc='lower right')

plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)

plt.plot(epochs_range, loss, label='Training Loss')

plt.plot(epochs_range, val_loss, label='Validation Loss')

plt.legend(loc='upper right')

plt.title('Training and Validation Loss')

#plt.show()

sunflower_url = "https://storage.googleapis.com/download.tensorflow.org/example_images/592px-Red_sunflower.jpg"

sunflower_path = tf.keras.utils.get_file('Red_sunflower', origin=sunflower_url)

img = keras.preprocessing.image.load_img(

sunflower_path, target_size=(img_height, img_width)

)

img_array = keras.preprocessing.image.img_to_array(img)

img_array = tf.expand_dims(img_array, 0) # Create a batch

model.save("model_flower.h5")

new_model = tf.keras.models.load_model('model_flower.h5')

predictions = new_model.predict(img_array)

score = tf.nn.softmax(predictions[0])

print(

"This image most likely belongs to {} with a {:.2f} percent confidence."

.format(class_names[np.argmax(score)], 100 * np.max(score))

)

os.remove(sunflower_path)

[/bg_collapse]

Sau đó chạy file bằng lệnh dưới là ra kết quả
python image_classification.py

Sử dụng hình ảnh bên dưới để dự đoán

Kết quả dự đoán là dandelion (hoa Bồ Công Anh) với 80% confident

Giải thích source code

Bước 1: import các thư viện cần thiết
Bước 2: load ảnh và chia thành 2 bộ: 80% ảnh cho training set và 20% ảnh cho bộ valid set
Bước 3: setup các tham số dùng để training
Bước 4: khởi tạo model Neural Network với nhiều layers
Bước 5: compile model
Bước 6: optimize model bằng Dropout để giảm thời gian xử lý
Bước 7: training model
Bước 8: predict hình ảnh

8 bước trên có các bước trước khi training khá phức tạp, các bạn đón đọc các bài tiếp theo để nắm được nguyên lý.

Sự khác biệt giữa code Python ở trên với code trong Notebook

Đoạn code ở trên đã được chỉnh sửa nhiều để có thể chạy trên Windows 10.

Không download ảnh trong code nữa, user download về sẵn
Thay thế hàm load files ảnh
Thêm code save model tại dòng: model.save(“model_flower.h5”)
Thêm code load model tại dòng: new_model = tf.keras.models.load_model(‘model_flower.h5’)
Thêm code xóa ảnh đã predict tại dòng: os.remove(sunflower_path)

Chúc các bạn thành công. Còn không thành công? Contact me with detail on the website footer.

Phân loại hình ảnh với Tensorflow – Bài 2 chạy code Python trên Windows 10

Chuẩn bị

Source code

Giải thích source code

Sự khác biệt giữa code Python ở trên với code trong Notebook

Leave a Reply Cancel reply