TensorFlow迁移学习DenseNet121预测10-monkey-species

 

In [1]:

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 [14]:

densenet121 = keras.applications.DenseNet121(include_top=False, pooling='avg', weights='imagenet')

 

In [3]:

for layer in densenet121.layers[0:-5]:
    layer.trainable = False

classes = 10
model = keras.models.Sequential([densenet121, keras.layers.Dense(classes, 'softmax')])
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['acc'])
model.summary()

 

 

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 densenet121 (Functional)    (None, 1024)              7037504                                                 
 dense (Dense)               (None, 10)                10250                                                 
=================================================================
Total params: 7047754 (26.89 MB)
Trainable params: 49162 (192.04 KB)
Non-trainable params: 6998592 (26.70 MB)
_________________________________________________________________

 

In [4]:

model.layers

 

Out[4]:

[<keras.src.engine.functional.Functional at 0x2b080120520>,
 <keras.src.layers.core.dense.Dense at 0x2b0801332e0>]

 

In [5]:

# 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 [6]:

# 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.densenet.preprocess_input,    # 使用原版densenet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.densenet.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 [7]:

# 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 [==============================] - 48s 1s/step - loss: 1.8848 - acc: 0.3865 - val_loss: 0.9695 - val_acc: 0.8281
Epoch 2/10
34/34 [==============================] - 44s 1s/step - loss: 0.9487 - acc: 0.7936 - val_loss: 0.4936 - val_acc: 0.9297
Epoch 3/10
34/34 [==============================] - 44s 1s/step - loss: 0.6071 - acc: 0.8715 - val_loss: 0.3262 - val_acc: 0.9531
Epoch 4/10
34/34 [==============================] - 45s 1s/step - loss: 0.4350 - acc: 0.9174 - val_loss: 0.2489 - val_acc: 0.9570
Epoch 5/10
34/34 [==============================] - 44s 1s/step - loss: 0.3377 - acc: 0.9437 - val_loss: 0.1934 - val_acc: 0.9688
Epoch 6/10
34/34 [==============================] - 45s 1s/step - loss: 0.3038 - acc: 0.9456 - val_loss: 0.1808 - val_acc: 0.9688
Epoch 7/10
34/34 [==============================] - 45s 1s/step - loss: 0.2753 - acc: 0.9447 - val_loss: 0.1394 - val_acc: 0.9727
Epoch 8/10
34/34 [==============================] - 44s 1s/step - loss: 0.2277 - acc: 0.9512 - val_loss: 0.1294 - val_acc: 0.9727
Epoch 9/10
34/34 [==============================] - 47s 1s/step - loss: 0.2135 - acc: 0.9568 - val_loss: 0.1096 - val_acc: 0.9805
Epoch 10/10
34/34 [==============================] - 46s 1s/step - loss: 0.1887 - acc: 0.9644 - val_loss: 0.1096 - val_acc: 0.9766

 

In [8]:

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

 

 

 

In [9]:

model.evaluate(valid_generator)

 

 

9/9 [==============================] - 9s 962ms/step - loss: 0.1125 - acc: 0.9743

Out[9]:

[0.11253786087036133, 0.9742646813392639]

 

In [10]:

monkey = plt.imread('./10-monkey-species/n7013.jpg')
plt.imshow(monkey)
monkey.shape

 

Out[10]:

(800, 1200, 3)

 

 

In [11]:

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

 

Out[11]:

(224, 224, 3)

 

In [12]:

monkey_zoomed = keras.applications.densenet.preprocess_input(monkey_zoomed)
monkey_zoomed = monkey_zoomed.reshape(1, width, height, 3)
predict = model.predict(monkey_zoomed)
predict

 

 

1/1 [==============================] - 1s 984ms/step

Out[12]:

array([[1.2785053e-03, 1.1586951e-03, 1.1033794e-03, 8.9743080e-05,
        6.6979276e-04, 6.0248189e-02, 1.6423552e-03, 9.3309772e-01,
        5.2002951e-04, 1.9154577e-04]], dtype=float32)

 

In [13]:

predict.argmax(axis=1)    # 预测正确

 

Out[13]:

array([7], dtype=int64)