- 导入库
import matplotlib.pyplot as plt
import seaborn as sns
import keras
from keras.layers import *
from keras.models import *
from sklearn.metrics import classification_report, confusion_matrix
from keras.preprocessing import image
from keras.callbacks import EarlyStopping
from keras.callbacks import ModelCheckpoint
from keras.callbacks import LearningRateScheduler
from keras.applications.densenet import DenseNet121
import numpy as np # linear algebra
import pandas as pd
- 目录+模型训练+测试
import os
for dirname, _, filenames in os.walk(r'Research Work\Data'):
for filename in filenames:
os.path.join(dirname, filename)
my_data_dir = r'Research Work\Data'
test_path = my_data_dir+'/test/'
train_path = my_data_dir+'/train/'
image_size = (224, 224,3)
batch_size=32,
train_datagen = image.ImageDataGenerator(
rescale = 1./255,
)
test_datagen = image.ImageDataGenerator(rescale = 1./255)
early_stop = EarlyStopping(monitor='val_loss',patience=100, verbose = 1)
batch_size = 32
train_generator = train_datagen.flow_from_directory(
train_path,
target_size = (224,224),
batch_size = 32,
class_mode = 'categorical')
from keras import backend as K
def recall_m(y_true, y_pred):
true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))
recall = true_positives / (possible_positives + K.epsilon())
return recall
def precision_m(y_true, y_pred):
true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))
precision = true_positives / (predicted_positives + K.epsilon())
return precision
def f1_m(y_true, y_pred):
precision = precision_m(y_true, y_pred)
recall = recall_m(y_true, y_pred)
return 2*((precision*recall)/(precision+recall+K.epsilon()))
train_generator.class_indices
test_image_gen = test_datagen.flow_from_directory(
test_path,
target_size = (224,224),
batch_size = 32,
class_mode = 'categorical')
class_name=test_image_gen.class_indices
epochs = 100
stepsperepoch=9
validationsteps=1
annealer = LearningRateScheduler(lambda x: 1e-3 * 0.95 ** x)
es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=100)
mc = ModelCheckpoint("own.h5", monitor='val_loss',save_best_only=True, mode='min',verbose=1)
input_t = Input(shape=(224, 224, 3))
model = DenseNet121(
include_top=True,
weights=None,
input_tensor=input_t,
input_shape=None,
pooling=None,
classes=3,
)
model.compile(loss='categorical_crossentropy',optimizer="RMSprop",metrics=['accuracy',f1_m,precision_m, recall_m])
model.summary()
hist = model.fit_generator(
train_generator,
epochs=epochs,
callbacks=[annealer,mc,es],
steps_per_epoch=stepsperepoch,
validation_data=test_image_gen,
validation_steps = validationsteps,
)
metrics=model.evaluate(test_image_gen)
print ("Validation Loss = " + str(metrics[0]))
print ("Validation Accuracy = " + str(metrics[1]))
print ("Validation F1 Score = " + str(metrics[2]))
print ("Validation Precision = " + str(metrics[3]))
predictions = np.argmax(model.predict(test_image_gen), axis=-1)
print(predictions)
print(f'Testing loss: {metrics[0]}')
print(f'Testing accuracy: {metrics[1]}')
- 打印分类报告准确率为96%，但在分类报告中显示为50%。它打印哪个值
print(classification_report(test_image_gen.classes, predictions))
print(confusion_matrix(test_image_gen.classes, predictions))
sns.heatmap(confusion_matrix(test_image_gen.classes, predictions), annot=True)
输出
precision recall f1-score support
0 0.06 0.05 0.06 111
1 0.24 0.25 0.24 301
2 0.66 0.66 0.66 838
accuracy 0.50 1250
macro average 0.32 0.32 0.32 1250
weighted aveg 0.51 0.50 0.51 1250