TensorFlow迁移学习Resnet50预测10-monkey-species

 

In [15]:

from tensorflow import keras
import tensorflow as tf
import numpy as np
import pandas as pd
from scipy import ndimage
import matplotlib.pyplot as plt

 

In [2]:

resnet50 = keras.applications.ResNet50(include_top=False, pooling='avg')

 

In [3]:

classes = 10
model = keras.models.Sequential()
model.add(resnet50)
model.add(keras.layers.Dense(classes, activation='softmax'))

 

In [4]:

model.layers

 

Out[4]:

[<keras.src.engine.functional.Functional at 0x25ed1b6fb80>,
 <keras.src.layers.core.dense.Dense at 0x25ed1b06220>]

 

In [5]:

# 把除最后一层之外的参数都冻结
model.layers[0].trainable = False
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['acc'])
model.summary()

 

 

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 resnet50 (Functional)       (None, 2048)              23587712                                                   
 dense (Dense)               (None, 10)                20490                                                               
=================================================================
Total params: 23608202 (90.06 MB)
Trainable params: 20490 (80.04 KB)
Non-trainable params: 23587712 (89.98 MB)
_________________________________________________________________

 

In [6]:

# 文件下载地址  https://www.kaggle.com/datasets/slothkong/10-monkey-species
train_dir = './10-monkey-species/training/training'
valid_dir = './10-monkey-species/validation/validation'
label_file = './10-monkey-species/monkey_labels.txt'
df = pd.read_csv(label_file, header=0)
df

	Label	Latin Name	Common Name	Train Images	Validation Images
0	n0	alouatta_palliata\t	mantled_howler	131	26
1	n1	erythrocebus_patas\t	patas_monkey	139	28
2	n2	cacajao_calvus\t	bald_uakari	137	27
3	n3	macaca_fuscata\t	japanese_macaque	152	30
4	n4	cebuella_pygmea\t	pygmy_marmoset	131	26
5	n5	cebus_capucinus\t	white_headed_capuchin	141	28
6	n6	mico_argentatus\t	silvery_marmoset	132	26
7	n7	saimiri_sciureus\t	common_squirrel_monkey	142	28
8	n8	aotus_nigriceps\t	black_headed_night_monkey	133	27
9	n9	trachypithecus_johnii	nilgiri_langur	132	26

 

In [7]:

# 图片数据生成器
height = 224
width = 224
channels = 3
batch_size = 32
classes = 10
train_datagen = keras.preprocessing.image.ImageDataGenerator(
    preprocessing_function = keras.applications.resnet50.preprocess_input,    # 使用原版resnet50自身的数据处理
    rotation_range = 40,   # 随机旋转 0~40°之间
    width_shift_range = 0.2, # 随机水平移动
    height_shift_range = 0.2, # 随机垂直移动
    shear_range = 0.2, # 随机裁剪比例
    zoom_range = 0.2, # 随机缩放比例
    horizontal_flip = True,  # 随机水平翻转
    vertical_flip = True, # 随机垂直翻转
    fill_mode = 'nearest', # 填充模式
)
train_generator = train_datagen.flow_from_directory(train_dir, target_size=(height, width),
                                                    batch_size=batch_size, shuffle=True, class_mode='categorical')

valid_datagen = keras.preprocessing.image.ImageDataGenerator(
    preprocessing_function = keras.applications.resnet50.preprocess_input
)
valid_generator = valid_datagen.flow_from_directory(valid_dir, target_size=(height, width),
                                                    batch_size=batch_size, shuffle=False, class_mode='categorical')

 

 

Found 1098 images belonging to 10 classes.
Found 272 images belonging to 10 classes.

 

In [8]:

# 训练
history = model.fit(train_generator, steps_per_epoch=train_generator.samples//batch_size, epochs=10,
                    validation_data=valid_generator, validation_steps=valid_generator.samples//batch_size)

 

 

Epoch 1/10
34/34 [==============================] - 41s 1s/step - loss: 1.4802 - acc: 0.5394 - val_loss: 0.4833 - val_acc: 0.9062
Epoch 2/10
34/34 [==============================] - 41s 1s/step - loss: 0.4419 - acc: 0.8856 - val_loss: 0.2153 - val_acc: 0.9531
Epoch 3/10
34/34 [==============================] - 41s 1s/step - loss: 0.2977 - acc: 0.9184 - val_loss: 0.1501 - val_acc: 0.9727
Epoch 4/10
34/34 [==============================] - 41s 1s/step - loss: 0.2131 - acc: 0.9531 - val_loss: 0.1265 - val_acc: 0.9805
Epoch 5/10
34/34 [==============================] - 42s 1s/step - loss: 0.1880 - acc: 0.9531 - val_loss: 0.0946 - val_acc: 0.9922
Epoch 6/10
34/34 [==============================] - 41s 1s/step - loss: 0.1616 - acc: 0.9662 - val_loss: 0.0941 - val_acc: 0.9883
Epoch 7/10
34/34 [==============================] - 41s 1s/step - loss: 0.1419 - acc: 0.9662 - val_loss: 0.1031 - val_acc: 0.9805
Epoch 8/10
34/34 [==============================] - 42s 1s/step - loss: 0.1120 - acc: 0.9709 - val_loss: 0.0717 - val_acc: 0.9922
Epoch 9/10
34/34 [==============================] - 39s 1s/step - loss: 0.1212 - acc: 0.9728 - val_loss: 0.0757 - val_acc: 0.9844
Epoch 10/10
34/34 [==============================] - 39s 1s/step - loss: 0.1045 - acc: 0.9700 - val_loss: 0.0707 - val_acc: 0.9844

 

In [14]:

pd.DataFrame(history.history).plot(figsize=(8, 5))
plt.grid()
plt.gca().set_ylim(0, 1.5)
plt.show()

 

 

 

In [43]:

model.evaluate(valid_generator)    # 准确率非常高

 

 

9/9 [==============================] - 7s 721ms/step - loss: 0.0705 - acc: 0.9853

Out[43]:

[0.07051555812358856, 0.9852941036224365]

 

In [53]:

monkey = plt.imread('./10-monkey-species/n9010.jpg')
plt.imshow(monkey)
monkey.shape

 

Out[53]:

(311, 472, 3)

 

 

In [54]:

x = width / monkey.shape[0]
y = height / monkey.shape[1]
monkey_zoomed = ndimage.zoom(monkey, (x, y, 1))
monkey_zoomed.shape

 

Out[54]:

(224, 224, 3)

 

In [55]:

monkey_zoomed = keras.applications.resnet50.preprocess_input(monkey_zoomed)
monkey_zoomed = monkey_zoomed.reshape(1, width, height, 3)   # 这里1就是每批次样本数
predict=model.predict(monkey_zoomed)predict

 

 

1/1 [==============================] - 0s 64ms/step

Out[55]:

array([[2.0728735e-03, 2.4864163e-02, 8.3858357e-04, 7.8938156e-03,
        2.2254630e-04, 1.1891891e-03, 1.4768102e-03, 1.1359336e-03,
        1.0765824e-03, 9.5922953e-01]], dtype=float32)

 

In [57]:

predict.argmax(axis=1)    # 预测正确

 

Out[57]:

array([9], dtype=int64)